M.Sc.

Graduation Project:

Soil-structure interaction modelling in performance-

based seismic jetty design

Final report

Simplified dynamic analysis

Uncoupled site and

Coupled 3D nonlinear site- structure dynamic analysis
structure dynamic analysis

Editor: Besseling, F. (Floris)

Graduation Committee: Prof. ir. A.W.C.M. Vrouwenvelder

(Chairman of the Committee, TU Delft, section Structural mechanics)
Prof. dr. A.V. Metrikine (TU Delft, section Structural Mechanics)
Dr. ir. R.B.J. Brinkgreve (TU Delft, section Geo-engineering & Plaxis)
Ir. H.J. Lengkeek (Internship supervisor, Witteveen+Bos)
Ir. L.J.M. Houben (TU Delft)
th
Delivery date: 19 of June 2012
 M.Sc. Graduation Project:

Soil-structure interaction modelling in performance-

based seismic jetty design

Editor: Besseling, F. (Floris)

Student number: 4025768

Email: f.besseling@student.tudelft.nl

Graduation Committee: Prof. ir. A.W.C.M. Vrouwenvelder

(Chairman of the Committee, TU Delft, section Structural Mechanics)
Prof. dr. A.V. Metrikine (TU Delft, section Structural Mechanics)
Dr. ir. R.B.J. Brinkgreve (TU Delft, section Geo-engineering & Plaxis)
Ir. H.J. Lengkeek (Internship supervisor, Witteveen+Bos)
Ir. L.J.M. Houben(TU Delft)
th
Delivery date: 19 of June 2012

Status: Final

Number of pages: 108

Summary
The importance of soil structure interaction in seismic design of structures is a recognized by the modern seis-
mic design community, that is very much moving towards performance based design principles. Particularly for
structures with deep foundations in soft soil conditions, soil-structure interaction is recognized to be an impor-
tant factor that has to be considered in design. Jetty structures obviously are such structures. On the contrary
seismic design standards do hardly provide any straight forward tools for engineers to account for soil-
structure interaction in design. It is clear that a problem exists, which has initiated this study.

In the first phase of the study a literature review is conducted, where important developments relating to per-
formance based seismic jetty design and soil-structure interaction are collected. Based on this literature review
three suitable design approaches are found for jetty structure design, being simplified dynamic analysis (push-
over + response spectrum), uncoupled dynamic analysis and coupled dynamic analysis.

Then the second phase of the study is geared towards static pile-soil interaction and pushover analysis of single
piles and jetty structures by modelling the soil as a conventional Winkler foundation or by performing more ad-
vanced hardening soil finite element analysis. Proper selection of clay parameters for the hardening soil model
hereby is a critical step in order to be able to verify p-y expressions. After calibration of Winkler foundations for
single piles, related issues like pile group effects are studied by means of finite element pushover analysis and
subsequently p-multipliers for these effects can be defined for dynamic jetty analysis.

Subsequently the focus is shifted towards free field dynamic analysis of soil deposit for vertically propagating
shear waves, because of its importance for seismic analysis of deep foundations. Both linear and nonlinear fi-
nite element modelling are performed and compared to equivalent linear frequency domain analysis solutions
for layered soils. The conclusions drawn from this phase are important input for the jetty dynamic analysis.

Then the seismic jetty response was calculated along the different methods. First by means of simplified dy-
namic analysis and then by both uncoupled and coupled dynamic analysis. Based on the results recommenda-
tions are proposed regarding the importance of soil-structure interaction for future jetty projects and how it
should be accounted for, where both performance as well practical issues are taken into consideration..
Table of contents
Summary .......................................................................................................................................................... 1
Table of contents ............................................................................................................................................. 1
1 Preface.................................................................................................................................................. 1
2 Introduction.......................................................................................................................................... 2
3 Project Description ............................................................................................................................... 3
4 Theoretical Background........................................................................................................................ 4
4.1 Earthquakes 4
4.1.1 Earthquake origin 4
4.1.2 Magnitude, intensity and seismic moment 4
4.1.3 Soil waves and attenuation 5
4.1.4 Site amplification effects 6
4.2 Seismic hazard assessment, probabilistic or deterministic approach 6
4.3 Performance based design 6
4.3.1 Structural importance classes 7
4.3.2 Earthquake motion levels 7
4.3.3 Performance objectives, ductility and damage criteria 9
4.4 Available seismic design methods 14
4.4.1 Geotechnical soil site response analysis 14
4.4.2 Structural seismic design methods 16
4.5 Jetty structural dynamics (under earthquake excitation) 19
4.5.1 Mathematical representation of systems under seismic loading 19
4.5.2 Coupling of superstructure, foundation and soil 20
4.5.3 Higher mode contributions in response 21
4.5.4 Damping 22
4.6 Soil representation in pile-soil interaction modelling 27
4.6.1 General comments on soil representation in seismic modelling 27
4.6.2 Linearization of soil response 27
4.6.3 The beam on Winkler-foundation concept and nonlinear p-y curves 28
4.6.4 Continuum material models for soil 39
5 Case study project .............................................................................................................................. 44
5.1 Jetty location and geometry 44
5.2 Seismic activity at the project location 44
5.3 Geotechnical characterisation 45
5.4 Seismic site characterization 46
5.4.1 Turkish/ISO seismic regulations response spectra 47
5.4.2 Eurocode EN-1998 (2005) response spectra 48
5.5 Jetty structural characteristics 48
6 Analysis method ................................................................................................................................. 51
6.1 Introduction 51
6.2 Pushover analysis 53
6.2.1 Static pushover analysis on conventional piles supported by Winkler foundation models 53
6.2.2 Static pushover analysis on a single pile embedded by a continuum with the Hardening
Soil constitutive models in Plaxis 54
6.2.3 Equivalent Plaxis 2d pile model based on the Plaxis 3d and Winkler model pushover
characteristics 55
6.2.4 Pushover analysis of the case study jetty cross-section in transverse direction 55
6.3 Simplified dynamic analysis based on the push-over results 56
6.4 Free field site response analysis 57
6.4.1 Input signals 58
6.4.2 Dynamic soil material parameters 59
6.4.3 Free field site response analysis by 1D Fourier frequency domain analysis 61
6.4.4 Free field site response analysis by Plaxis 2D/3D finite element analysis 61
6.5 Dynamic analysis of jetty on nonlinear Winkler foundation for support node motions resulting
from free field analysis 63
6.6 Coupled transverse dynamic analysis of a slice soil deposit and jetty structure 64
7 Results ................................................................................................................................................ 65
7.1 Nonlinear Pushover Analysis 65
7.1.1 Pushover Analysis on 2D wall models compared to Winkler pile models 65
7.1.2 Pushover analysis on 3D single pile models 67
7.1.3 Linearization of pile-soil interaction 74
7.1.4 Equivalent Plaxis 2D model to represent lateral single pile response 74
7.1.5 Pushover analysis of jetty cross-section (3 rows of piles in loading direction) 76
7.2 Free field site response analysis 82
7.2.1 Calibration of the finite element model 82
7.2.2 Comparison of responses from the linear elastic finite element model and equivalent
linear frequency domain analysis for layered soil profile 84
7.2.3 Site response for linear elastic, Mohr-Coulomb, HS and HSsmall soil constitutive models 86
7.2.4 Operation Plaxis 3D dynamics module (Release 2011.01 February 2012) 93
7.3 Jetty simplified dynamic analysis according to Fajfar N2-response spectrum method 94
7.4 Uncoupled and coupled site and jetty nonlinear dynamic time domain analysis 97
7.4.1 Summary of acceleration, displacement and pile head bending moment response
quantities 97
7.4.2 Displacement time histories and shifts of system equilibrium position 100
7.4.3 Kinematic pile loading 102
7.4.4 Conclusions dynamic jetty analysis 103
7.5 Conclusions regarding jetty transverse seismic analysis 104
8 Conclusions and Discussion .............................................................................................................. 106
9 Suggestions for further research ...................................................................................................... 108
Appendices....................................................................................................................................................... 1
References ....................................................................................................................................................... 3
1 Preface
This master thesis finalizes my master study Structural Engineering, specialization Structural Mechanics at Delft
University of Technology. The project as part of the master track covers 40 ECTS and was executed in the pe-
riod ranging from September 2011 to June 2012.

Over this period I have almost continuously enjoyed the interesting project. Now the master thesis scope has
finished, but still interesting questions relating the soil-structure interaction problem arise. I definitely will try
to address some of these during the remainder of my career in structural/geotechnical engineering.

The study, titled “Soil structure interaction modelling in seismic jetty design” is commissioned by Wit-
teveen+Bos Consulting Engineers and was executed at Witteveen+Bos Deventer under the supervision of Ir.
H.J. (Arny) Lengkeek. I would like to express my gratitude to Witteveen+Bos for offering me this position and a
special gratitude to Arny for his dedicated support.

The graduation committee for this project consist of the following members:
Prof. ir. A.W.C.M. Vrouwenvelder (Chairman of the Committee, TU Delft, section Structural Mechanics)
Prof. dr. A.V. Metrikine (TU Delft, section Structural Mechanics)
Dr. ir. R.B.J. Brinkgreve (TU Delft, section Geo-engineering & Plaxis)
Ir. H.J. Lengkeek (Internship supervisor, Witteveen+Bos)
Ir. L.J.M. Houben (TU Delft)
I would like to thank the committee for their guidance and support, and additionally the Plaxis technical sup-
port team for solving many problems together.

Special thanks to Lotte for her great support and to my parents providing me the opportunity to finish my mas-
ter study in Delft.
th
Deventer, 19 of June 2012

Floris Besseling

All rights reserved. No part of this publication including appendices may be reproduced, without the prior per-
mission of the author and under condition of full reference. Contact: f.besseling@witteveenbos.nl

1
2 Introduction

Earthquakes are among the most devastating natural disasters humans have faced over history. Since civilisa-
tion has developed, and demand for all kind of buildings and other type of structures has increased, the risk re-
lated has increased inherently. Even over the past few years direct and indirect effects by earthquake result
thousands of fatalities when affecting densely populated areas like in Southern Sumatra - Indonesia (2009),
Haiti Region (2010), Southern Qinghai - China (2010) and Japan (2011), with 1,117, 316,000, 2,968 and 20,352
fatalities respectively.

With the development of civilisation during the last century, buildings and, infrastructure have increased expo-
nentially in number and size, which inherently has increased the risks related to earthquakes. In this perspec-
tive ports are of special importance because of their major significance for earthquake recovery and the large
amount of hazardous materials that are transferred and stored in these areas. Hence, earthquake resistance
and safety of port structure needs special attention from this perspective.
Port structures may be subdivided in several categories like gravity quay walls, sheet pile quay walls, pile sup-
ported wharves and piers/jetties, cellular quay walls, quay walls with cranes and breakwaters. This thesis fo-
cuses primarily on pile supported piers, often referred to as jetties, piers or pile-deck systems. Typically, these
jetty structures consist of a concrete deck on top of series of long reinforced or prestressed concrete piles or
tubular steel piles. Over the past decades vessel sizes have increased resulting an increase in jetty size as well.
This results a global trend switch to the use of tubular steel piles, as was already common in offshore engineer-
ing practice and hence is taken as a starting point in this study.

Traditionally, soil-structure interaction effects were ignored in seismic design of structures, since they were be-
lieved to only have favourable effects. The lengthening of the period shifts the structure response to the spec-
tral branch of lower accelerations which implies a reduction of inertia forces in the structure. However along
modern performance based design principles soil-structure interaction effects are recognized to not necessarily
have beneficial but even may have very detrimental effects for the response of the superstructure (Priestley &
Park, 1987) (Mylonakis, Nikolaou, & Gazetas, 1997), (Mylonakis & Gazetas, 2000). The global trend shift to-
wards performance based design in the seismic engineering branch implies an increasing focus on displace-
ments rather than on inertia forces, which makes proper consideration of soil structure interaction a critical
factor. Additionally the failure of foundations their selves and possible effects of soil failure have become a
more important issue in seismic design. The effects of soil structure interaction have been subjective to re-
search for about half a century, but are still under discussion.

Code provisions relating to soil-structure interaction nowadays are still very limited and straight forward pro-
cedures to account for soil structure interaction in design are not included in most codes. Simplified dynamic
analysis methods are commonly used as a starting point, where the role and possible effects of soil-structure
interaction in the response often remain unclear. On the contrary research activities at universities all over the
world already are far ahead, providing a variety of knowledge in this field. The connection with engineering
practice however appears to be somehow lost, which forms a problem for practicing engineers dealing with de-
sign of structures in seismically active areas. This study focuses on the performance of available and usable de-
sign methods for practicing engineers that need to account for soil structure interaction in seismic jetty design.

The present study is concentrated at a case jetty project, that typically meets the conditions regarding deep
piled end-bearing foundations in soft soil, for which according to standards and literature, soil-structure inter-
action effects are considered most important. Although the case project typically represents the common
situation of deep end-bearing jetty foundations in a soft soil deposit, it should be noted that differently situ-
ated jetty structures may result different critical design issues. However, since master’s thesis research re-
quirements limits the scope of the research, it is chosen to focus primarily on this typical situation, not per-
forming a complete parametric study on varying local conditions. This latter study is recommended for further
study by a future graduating student or may be attended as part of future Witteveen+Bos jetty projects in
seismic areas.

2
3 Project Description

This study exists of four main phases:

The first part is a literature study of seismic engineering principles and related geotechnical as well structural
aspects, with a focus on soil structure interaction and analysis of structures with deep foundations in soft soil
conditions. The literature study is based on research papers, seismic design codes and seismic engineering text
books. First, seismic design principles in general are studied. Recommendations regarding how to design struc-
tures for seismic loading will be summarized and conclusions will be drawn regarding the most suitable seismic
design methods for jetty structures. Subsequently the focus is shifted more towards typical jetty structure re-
lated analysis issues, where soil structure interaction under seismic loading is in various literature and seismic
standards recognized to be a significant factor to be included in design.

The second part focuses on the static analysis of pile-soil interaction and laterally loaded piles A comparative
study is performed, in which the widely used beam on nonlinear Winkler foundation technique for laterally
loaded piles is verified by pushover analysis with geotechnical engineering finite element software Plaxis 3D,
making use of advanced soil constitutive models. Proper selection of hardening soil parameters herein was a
critical step in order to reasonably use the numerical modelling as a tool for p-y curve validation. Possibilities
for defining equivalent 2D models that properly represent the horizontal jetty response are investigated. Based
on finite element analysis of pile groups p-multipliers are established that correctly reduce Winkler foundation
capacity for pile group effects.

Then the third phase of the study addresses free field site response analysis for a soil deposit overlying the
stiffer bedrock, subjected to vertically propagating shear waves. This common step in earthquake geotechnical
engineering provided soil deposit and soil surface motions for selected bedrock input signals. Subsequently in
simplified dynamic analysis procedures usually the soil surface acceleration amplitude is applied in a response
spectrum procedure. However when soil-structure interaction by the pile shafts is to be accounted for, the soil
deposit motion along the entire shaft has to be calculated. The free field responses are obtained by equivalent
linear frequency domain analysis and Plaxis 2D analysis, where ultimately the performance of the very new
Plaxis 3D dynamic module was verified. Typical performances of different available Plaxis soil constitutive mod-
els in dynamics are studied and recommendations regarding their application are formulated.

Subsequently in the fourth phase of the study simplified dynamic analysis and uncoupled and coupled dynamic
analysis are performed for the site and structure. Based on the jetty lateral load-deformation characteristics
obtained from pushover analysis, simplified dynamic analysis will be performed. The calibrated green field soil
models resulting from the previous stage are used as a basis for constructing coupled dynamic models and the
free field responses are applied to the structural model in uncoupled dynamic jetty analysis. The goal of this
stage is to compare and assess structure responses obtain along the different approaches and investigated how
soil structure interaction effects may affect the response

Finally, concluding remarks relating recommended seismic design procedures for jetty structures will be postu-
lated, where the importance of coupling of soil and structure response is considered as well.

3
4 Theoretical Background

4.1 Earthquakes

4.1.1 Earthquake origin

Earthquakes are caused by vibrations of the Earth’s surface due to a sudden release of energy somewhere in
the Earth’s crust. This energy is generated by a gradually build up of stresses along plate boundaries, induced
by continental drift of the tectonic plates, as can be described by the elastic rebound theory. The continental
drift is nowadays believed to be driven by convection currents within the Earth’s mantle forming a continuous
cycle of heating and cooling of the material it consists of.
Earthquakes can be characterized by various criteria. First of all they can be subdivided into two main catego-
ries, the inter-plate and intra-plate earthquakes. Intra-plate earthquakes are generate by faults in the plate in-
terior where inter-plate earthquakes are generated by faults along the plate boundaries. Generally the strong-
est earthquakes are generated by inter-plate faults, since build-up stresses are typically higher at these loca-
tions.
Plate boundaries are categorized as divergent, convergent or transcurrent, depending on the relative move-
ment of the adjacent plates. Additionally, three types of fault movements can be identified, i.e. the normal
fault, the reverse of thrust fault and the strike-slip fault as indicated in Figure 4.1.

Figure 4.1: Interplate boundary types (after Clough and Penzien (1993))

The nature of faults and their activity is accounted for in probabilistic seismic hazard assessments and there-
fore is of interest in earthquake engineering. Additionally the distance of the site considered to the fault is of
importance, where for near-fault projects vertical ground motions, directivity effects and high frequency con-
tent of the seismic signal are important issues to consider in design.

4.1.2 Magnitude, intensity and seismic moment

Earthquake levels can be described along different types of measuring scales. Important measures in earth-
quake engineering that sometimes are mixed up are magnitude, intensity and moment. Magnitude is the
amount of energy released from the earthquake source. Intensity represents the impact of an earthquake on a
specific site and so depends on distance from the epicentre and local soil conditions. Seismic moment is a
measure for the overall deformation at the fault and can be interpreted simply in terms of ground deformation.
Seismic moment is considered to be rather accurate in calculating the earthquakes source strength.
The most common magnitude scale applied in earthquake engineering practice is the Richter Scale, ranging
from 0 to 10. An intensity scale often used is the Modified Mercalli Intensity Scale (MMI) ranging form I to XII.

4
Seismic moment scales also have been proposed by researchers, e.g. (Hanks & Kanamori, 1979) and (Kanamori,
1983). More information regarding this earthquake measuring scales can be found in most earthquake engi-
neering textbooks.

4.1.3 Soil waves and attenuation

Once a tremor has been generated by a sudden release of energy, stress waves start propagating through the
Earth’s crust, causing soil deformations, which can be described along the wave propagation through solids
theory, for which one is referred to (Kramer, 1996; Verruijt, 2005).

Considering the soil to react to local disturbances as an elastic solid, a variety of wave types are allowed which
makes the resulting ground motion quite complex to determine.
First of all the waves can be subdivided into body waves and surface waves. Body waves again can be subdi-
vided into primary waves (also denoted as pressure waves or P-waves) and secondary waves (also denoted as
shear waves or S-waves). Shear waves can again be subdivided into Sh-waves and Sv-waves, indicating the direc-
tion of ground movement to be respectively horizontal or vertical. Elastic wave theory provides that P-waves
travel faster than S-waves and therefore will be the first waves to arrive. P-waves can travel through solids as
well liquids, where S-waves can travel through solids only. This implies that S-waves cannot travel through the
Earth’s liquid core where P-waves can.

Figure 4.2: Top left: Compression body wave, Bottom left: Shear body wave, Top right: Surface Rayleigh wave, Bottom
right: Surface Love wave (From S.L. Kramer, Geotechnical Earthquake Engineering (1996))

Surface waves also can be split up in two important sub-categories, i.e. Rayleigh waves and Love Waves.
Rayleigh waves are the result of interaction between P and Sv waves and their existence was first demonstrated
by the English physicist John William Strutt (Lord Rayleigh) in 1885. Rayleigh waves are analogous to water
waves and can be shown to exist in a homogeneous elastic half-space. Love waves are the result of interaction
between P and Sh waves and where deduced by the Britisch mathematician A.E.H. Love. Love waves require a
surficial layer of lower S-wave velocity compared to the underlying half-space. Other types of surface waves ex-
ist, but are considered to be much less relevant from an earthquake engineering point of view.

Wave propagation affect seismic signal while it travels away from its source. Thereby understanding and using
the principles of wave propagation make seismologist able to predict site soil response based on ground condi-
tions along the travel path of the wave. Principles of dispersion, refraction, diffraction and filtering of wave sig-
nals are of importance in this sense. Traditionally, seismologists accounted for wave propagation by empirical
local attenuation laws, which are averages found from recorded accelerograms. With increasing experience in
this field specialist become able to define local seismic signals more accurately, by making use of advanced
models. Nevertheless it should be noted that it is always to be recommended to check results by advanced
models against these local empirical attenuation laws and simple site response analysis techniques.

5
4.1.4 Site amplification effects
Fault characteristics together with wave propagation effects determine the local bedrock ground motion inten-
sity, frequency content and duration. Then to determine design ground level ground motions is to account for
site amplification effects. It is noted that in this study the effects of surface waves are not considered, where it
is restricted to the common engineering approach representing the seismic load by vertically propagating shear
waves only.

The site amplification effects are characterized by the natural frequencies of the soil deposit, which can be es-
timated (Kramer, 1996) based on the averaged present soil shear wave velocity and soil deposit height as:
vs  n
fn = 1 +  , n = 0,1, 2,.... (4.1)
4H  2
Often the soil deposit parameters are taken over the top 30 meters of the deposit, but whether this common
depth is appropriate is very much project dependent. Generally the top 30 meter approach is most appropriate
for rather short period content of the bedrock signal, as for longer periods seismic wave lengths are much
longer than 30 meters and response is likely to be affected by soil characteristics at much greater depths. From
equation (4.1) the fundamental frequency and fundamental period of the soil deposit are easily derived to be:
1 v
f0 = = s (4.2)
T0 4 H
It is noted that averaging the soil deposit characteristics over depth may result in inaccurate predictions of ac-
tual site amplification. A better estimate generally is obtained by performing site response analysis including
the layering of the deposit, for which one is referred 4.4.1.

4.2 Seismic hazard assessment, probabilistic or deterministic approach

In the preceding paragraphs, factors affecting a seismic site input signal were discussed. Ground motions can
be estimated on the basis of these factors, which is called site seismic hazard analysis. Two seismic hazard
analysis approaches are distinguished; i.e. deterministic seismic hazard assessment and probabilistic seismic
hazard assessment. In the first approach a particular governing earthquake scenario is deterministically as-
sumed to be governing based on the available information on seismic sources in the surrounding area. In the
second approach uncertainties in fault location and activity, earthquake magnitude, wave attenuation and re-
turn period are taken into account in a probabilistic calculation resulting in a local ground shaking level related
to a certain probability of exceedance. Generally for a specific construction site, different bedrock shaking lev-
els with related return period or probability of occurrence in a certain time interval are provided.
Modern seismic standards define ground motion levels mostly in the probabilistic setting. A remarkable excep-
tion in this sense is the new approach adopted in the U.S. NEHRP 2012 seismic design regulations, that for very
high seismicity areas shift from probabilistic-based ground motion levels to maximum-to-be-expected ground
motion levels for near active fault areas in order to obtain more economic designs.

Probabilistic seismic hazard assessment procedures are subdivided by many standards in a simplified code
based approach and a detailed site specific probabilistic seismic hazard assessment. The simplified code based
method defines design peak accelerations based on topographic location and local soil conditions only, where
in the site specific detailed assessment procedure all possible influencing factors may be accounted for. For
sites where strong ground motions are to be expected or soil conditions are very much varying or weak, a de-
tailed site specific seismic hazard assessment is recommended by most standards. Estimating ground motion
levels and its reliance on incomplete or uncertain information makes the seismic hazard analysis process de-
pendent on unavoidable subjective decisions by the engineer. Hence it is noted that an additional check of the
resulting ground motions by simplified assessment procedures is recommended as well.

4.3 Performance based design

In traditional structural engineering and earthquake engineering practice most structures were designed pri-
marily based on static, or equivalent static loads (derived from inertia forces). However, the conventional force
based approach has proved to suffer from several shortcomings when used in seismic design, which are related
to the extreme load levels for high intensity earthquakes. These shortcomings are mostly related to:

6
- Interdependency of strength and stiffness;
- Uncertainties related to appropriate force reduction factors and ductility capacity
- Redistribution possibilities in irregular structures when properly designed for.
Generally too heavy designed structures result from purely force based design, for which it seems hard to en-
sure sufficient ductility under earthquakes exceeding the design level earthquake. Furthermore also redistribu-
tion of forces over structural elements is more difficult to appropriately account for in structural design and a
clearer view on possible failure modes may result more effective and consequently more economic designs.
Since lessons learned after strong earthquakes in the 1990s a trend switch has taken place in which perform-
ance based design has become an increasingly popular methodology in earthquake engineering practice.

The key concept in performance/deformation based design is the inclusion of design damage or deformation
criteria rather than design forces as a starting point in seismic design. Post-elastic structural behaviour is pref-
erably directly incorporated in design were coupling with ductility and damping can be achieved. Generally, the
design deformation objectives are related to a structural importance class, combined with a multi-level (usually
a dual-level) intensity approach. Important details relating to this two-way approach are presented in the next
sections for the following seismic regulations:
- Eurocode Regulations for Earthquake Resistant Design (NEN-EN 1998, 2004)
- Turkish Technical Seismic Regulations for Port and Harbour Structures (Turkish Technical Seismic
Regulations, 2008)
- ISO Seismic Design Regulations (ISO 19901-2, 2004; ISO 19902:2006, 2006; ISO 19903, 2006)
- U.S. NEHRP and MOTEMS Seismic Design Regulations (NEHRP Fema-450, 2012; MOTEMS, 2010)
- PIANC Seismic Design Guidelines for Port Structures (PIANC, 2001)

Simplified performance based seismic design techniques for buildings have been under discussion already for
many years and are therefore widely available. However, for port structures accuracy of simplified methods is
less clear. Goel has proposed performance based design methods for pile-deck systems (Goel, 2010), however
it is noted that accuracy of the analytically derived expressions is not verified by experiments which will be
done in the near future.

4.3.1 Structural importance classes

Seismic standards generally specify performance objectives based on a structure importance class and an
earthquake motion level. The former is presented herein as it is described in Eurocode (importance classes I, II,
III or IV). Other seismic codes have somewhat different nomenclature, MOTEMS defines a Risk Classification
(High, Medium or Low), the Turkish Seismic Design Regulations define a structure category (Special, Normal,
Simple, Unimportant), NEHRP defines a Seismic Use Groups (SUG I, II and III) and PIANC follows the Japanese
Standard that defines structure classes (Special, A, B or C) Main properties of port structures that are decisive
for the related importance class are:
- If the structure is open to the public
- To which extended is the structure of importance for earthquake recovery
- If hazardous materials will be treated at the structure (Eurocode, Turkish Reg., NEHRP, PIANC)
- Maximum vessel size and annual amount of cargo transferred at the structure (MOTEMS)
In the following paragraph earthquake motion levels as defined in various seismic standards are summarized,
based on structure importance class IV or equivalent.

4.3.2 Earthquake motion levels

In the common dual-level design approach, two design earthquake motion levels are defined and related per-
formance objectives based on structure importance class are established.

7
Table 4.1: Dual level Earthquake motions for High Importance Class structures (including LNG/Oil Jetties) sum-
marizes earthquake motion levels provide by different seismic standards.

Table 4.1: Dual level Earthquake motions for High Importance Class structures (including LNG/Oil Jetties)
Standard Structure class Earthquake motion level
Probability Of Ex-
ceedance over period T
(POE) and mean return
period (MRP)
Eurocode 8 Importance class IV * Damage limitation re- POE = 17% in 50 years
quirement earthquake MRP = 261 years
No collapse requirement POE = 4% in 50 years
earthquake MRP = 1303 years
Turkish Seismic Regulations Special Structures E1 earthquake POE = 50% in 50 years
2008 MRP = 72 years
E2 earthquake POE = 10% in 50 years
MRP = 475 years
E3 earthquake POE = 2% in 50 years
MRP = 2475 years
NEHRP - Fema 450 Seismic Use Group III Maximum probable POE = 50% in 50 years
*** event (MPE) MRP = 72 years
Maximum reasonable to POE = 2% in 50 years
be expected earthquake MRP = 2500 years or de-
level terministically deter-
mined **
MOTEMS 2010 California High Risk Classification Level 1 earthquake POE = 50% in 50 years
Building Code MRP = 72 years
Level 2 earthquake POE = 10% in 50 years
MRP = 475 years ****
ISO 19901-2 Exposure level L1, Seis- Extreme level earthquake Undefined, because ELE
(simplified procedure) mic risk category 4 level related to ALE level
by reserve capacity coef-
ficient Cr. Back-
calculation by EC8 annual
Pexc expressions and Cr =
2 results: POE = 22% in
50 years, MRP = 312
years
Abnormal level earth- POE = 2% in 50 years
quake MRP = 2500 years
Pianc Special Class Level 1 earthquake likely POE = 50% in approxi-
to occur during structure mately 50 years, MRP =
life span 72 years
Level 2 rare infrequent POE = 10% in approxi-
high intensity earthquake mately 50 years, MRP =
475 years
* According to Eurocode 8 ground peak accelerations are multiplied by an importance factor. In order to then
obtain equivalent motion levels a formula is suggested to calculated annual rate of exceedance probabilities
corresponding to different motion levels: H (ag ) = k0 (ag ,ref γ 1 ) − k , where k0 is a site specific constant, γ 1 is
the importance factor and k is a constant approximately equal to 3.
* In FEMA-450 Commentary 1.2 the related earthquake level is termed the reasonably representative of the
most severe ground motion for moderate seismic zones. Subsequently one design structures for this maximum
considered earthquake multiplied with factor 2/3 to the performance criteria applicable to the damage control
limit state. Additionally it is stated that for zones of high seismicity a PSHA may give too high structural de-
mands, which are much higher than structural demands for characteristic magnitudes of earthquakes gener-

8
ated every few hundred years by these faults. Therefore a deterministic procedure is proposed for calculation
near fault high seismicity ground motion levels.
** NEHRP couples SUG to response quantities and performance objectives, not directly to ground motion in-
tensity.
*** In the new ASCE standard currently under development, which probably will also affect future MOTEMS
standards, level 2 earthquakes will be related to 2% in 50 years earthquakes.

As shown in Table 4.1, design earthquake levels vary considerably with the different seismic design standards.
Interestingly, the seismic port design standards MOTEMS (U.S.) and Pianc (a lot of input from the U.S.) provide
significantly lower return periods for the highest level design earthquake, compared to the Eurocode seismic
regulations and Turkish seismic regulations for Construction of Coastal and Harbour Structures.

4.3.3 Performance objectives, ductility and damage criteria

Seismic design standards generally provide performance objectives related to predefined safety targets, based
on structural classes and design earthquake levels. Additionally these objectives may be modified by the client
in consultation with regulators, or to meet regional requirements where they exist. The level to which per-
formance objectives are specified in different seismic standards differs considerably. Some codes exactly quan-
tify allowable material strains and structure displacements for different earthquake levels (Turkish Seismic
regulations, MOTEMS, Pianc), where others only generally specify a state of allowable damage (Eurocode 8, ISO
Seismic regulations).

In this research on seismic jetty design, the focus will be on performance objectives for these type of port
structures. It has to be noted that performance objectives for these type of structures are less common than
the detailed requirements formulated for building structures.

4.3.3.1 Eurocode 8
Eurocode 8 seismic regulations (NEN-EN 1998, 2004) do not provide specific performance objectives for jetties
or other type of pile supported port structures. General deformation requirements given in Eurocode 8 primar-
ily focus on verification of rotation in potential plastic hinges in general for any type of structue. These re-
quirements are provided in Eurocode 8 -1 and are referred to in other parts of the standard, including the part
on foundation design.

In order to be allowed to benefit of plastic hinging in seismic design, according to Eurocode 3, steel circular hol-
low sections need to be of section class 1 or 2, which is guaranteed if:
D 235
< 70 (4.3)
t fy
Where D and t are the tube diameter and thickness respectively and f y [MPa] is the material yield strength.
With respect to the two typical limit states (Damage Limitation Requirement and No Collapse Requirement),
Eurocode 8 requires no or very minor damage due to ground motions associated with the first limit state, how-
ever no typical limit rotations for piles are presented. With respect to the second limit state the code allows ro-
tations up to the real ultimate rotation capacity, divided by structural safety factor depending on kind of mate-
rial considered. It is stated that material related partial factors should account for strength degradation due to
cyclic loading. If not available, partial factor γM for persistent or transient loading may be applied.

According to Eurocode 8, piles should be designed to remain elastic, but may under certain conditions (that are
not further defined) be allowed to develop plastic hinge at their heads.

4.3.3.2 Turkish Technical Seismic Regulations (2008)

This paragraph on Turkish Technical Seismic Regulations performance objectives is based on the chapter on
seismic design of pile quays and jetties (Turkish Technical Seismic Regulations, 2008). The modern Turkish
Seismic Regulations are already strongly adapted to deformation based design principles, significantly further
than many other seismic design standards. The displacement based design is in these regulations fully inte-
grated and also appointed as such. Furthermore typically a complete chapter is related to seismic design of

9
port structures, where other seismic standards often lack specific regulations for this group of structures. Per-
formance objectives for jetty structures presented in these regulations are:
- Conform these seismic regulations, plastic hinges are not allowed to develop in decks of pile quays and
jetties. Plastic hinges may only develop in piles or monolithic pile-deck connections.
- For both the resistance-based design approach and the deformation-based design approach limit rela-
tive deformations at jetty deck level are presented:
- Limit strains for different earthquake levels are presented as shown in Table 4.2

Table 4.2: Limit strains for structural steel according to Turkish Seismic Regulations (2008)
Minimum Damage Controlled Damage
Resistance-based design 0.008 0.015
Deformation-based design 0.008 0.025

It is worth noting that the allowed material strains for the minimum damage performance objectives are con-
siderably beyond initial yielding, which seems to be a general trend in different seismic regulations. Apparently
much higher strains can be accepted according to this standard, even for the minimum damage requirements.

The pile curvatures corresponding to the limit strains shown in Table 4.2 may be calculated according to:
α ε ls
ϕls = (4.4)
D
Where α = 2.25 and α = 2.10 are dimensionless factors for circular concrete piles and rectangular concrete
piles respectively. For steel pipe piles no limit curvature expressions are given. To prevent for plate buckling
under these high stress/strain levels, the regulations require minimum t/D ratio’s for tubular steel piles to be:
D E
≤ c ≤ 80 (4.5)
t fy
Where c is a dimensionless factor ranging from 0.08 to 0.14, depending on the pile installation procedure and
eventual pile fill.

4.3.3.3 MOTEMS/NEHRP Fema 450

MOTEMS is the Californian seismic design guideline for marine oil terminals (MOTEMS, 2010). MOTEMS is part
of the California Building Code, is to a far extent adopted to the concept of performance-based design. Califor-
nian oil terminals that are subjective to high near fault seismicity are being designed according to these stan-
dards. The MOTEMS guidelines therefore partly overrule NEHRP Fema-450 when it comes to seismic resistant
of port structures in the U.S. Besides, NEHRP is primarily focussed on earthquake resistant design of buildings.
Hence in the present study performance objectives according to MOTEMS are treated, and performance objec-
tives according to NEHRP Fema-450 are not.

According to MOTEMS, plastic deformation mechanisms are not allowed in decks. Strain limits for steel pipe
piles as specified in MOTEMS are presented in Table 4.3.

Table 4.3Limit strains for structural steel according to MOTEMS (2010)

Level 1 Level 2
Limit strain εls at pile-deck connec- 0.008 0.025
tions
Limit strain εls at in ground plastic 0.004 0.008
hinges

Additionally it is stated that each structural component expected to undergo inelastic deformation shall be de-
fined by its moment-curvature relation. These moment curvature relations may be developed based on the
Level 1 and Level 2 limit strains as shown above, which can be translated into typical limit curvatures for steel
piles as:

10
 - Steel section first yield curvature:
2ε y
ϕy = (4.6)
D
- Steel section effective yield curvature:
(2.1 ~ 2.2)ε y ,eff
ϕ y ,eff = (4.7)
D
- Steel section limit strain curvature:
2ε ls
ϕls = (4.8)
D
Based on these limit strains, global structural displacement capacity may be determined by a nonlinear static
procedure. Yield curvatures and estimated plastic hinge length may subsequently be used to calculated rota-
tions and displacements at yield.

4.3.3.4 ISO Standards

ISO guidelines (ISO 19901-2, 2004) provide general descriptive performance requirements for both ELE (Ex-
treme Level Earthquake) and ALE (Abnormal Level Earthquake) design levels. However, no clear numerical limit
values for deformation parameters are provided.

Among the most important requirements for ELE design are:

- Limited nonlinear behaviour is permitted, brittle degradation shall be avoided
- Secondary structural components shall follow the same ELE design rigour as that of primary compo-
nents
- Joints strengths shall exceed calculated elastic joint forces and moments
- At foundation component element level an adequate margin shall exist with respect to axial and lat-
eral failure of piles. At the foundation system level an adequate margin shall exist with respect to
large-deflection mechanisms which would damage or degrade and require repairs to the structure or
its ancillary systems
- Not any loss of functionality in safety, escape or evacuation systems is allowed due to the ELE

With respect to ALE events the requirements are:

- Structural elements are allowed to exhibit plastic degrading behaviour, but catastrophic failures
should be avoided
- Stable plastic mechanisms in foundations are allowed, but catastrophic failure modes such as instabil-
ity or collapse should be avoided.
- Joints are allowed to exhibit limited plastic behaviour within their ultimate strengths, but shall be de-
signed to demonstrate sufficient ductility and residual strength at these deformation levels
- Safety, escape and evacuation systems shall remain functional during and after an ALE event

It may be clear that ISO performance objectives are very generally formulated, giving quite some freedom to
the engineer, while on the other hand covering all most important performance requirements for the dual level
approach.

4.3.3.5 Pianc
Assuming the jetty structure to be a Special Class structure, PIANC design guidelines for port structures (PIANC,
2001) require the structure to remain serviceable for a Level 1 and serviceable/repairable for a Level 2 earth-
quake event.

11
 Figure 4.3: Schematic figure of performance grades (From Pianc (2006))

Limit residual displacements for pile supported wharves, related to the Degree II damage level as provided by
PIANC are presented in Table 4.4. It should be noted that performance objectives as indicated in PIANC strictly
apply to wharves. It is reasonable that similar values apply to jetties, where it might be expected that perform-
ance objectives in transverse direction are less strict for jetties, but this will depend very much on use of the
jetty (chemicals, spill risks, etc.)

Table 4.4: Limit residual displacements according to Pianc (2006)

Degree II damage level
Differential settlement between deck and land be- 0.1~0.3 m
hind it
Residual tilting towards the sea 2° = π/90 rad

For piles and other structural components it is required that flexural failure should precede shear failure. PIANC
follows MOTEMS in the definition of allowable deformations related to the dual limit state approach and mo-
ment curvature relations. Again it may be noted that again the limit strain for the low level L1 earthquake
(0.008) exceeds the steel strain at yield by a factor 4 - 6, depending on the steel quality. This implies that ex-
periences from the past have shown that only minor damage will result for this relatively high limit strain level
that exceeds the strain at initial yielding by far. Limit strains subsequently can by linked to limit curvatures,
which then determine plastic rotation capacity as:
θ p = ϕ p L p = (ϕls − ϕ y ) L p (4.9)
Where the plastic hinge length ( L p ) depends on the section shape and the slope of the moment diagram in
the vicinity of the plastic hinge. Plastic hinge lengths in piles located below ground level are consequently also
affected by subgrade reaction stiffness of the soil, but since they are not repairable after a seismic event they
are generally not accepted.

PIANC provides that for structural analysis it will often be adequate to represent the response by a bi-linear
approximation as shown in Figure 4.4. It is noted that the representative yielding bending moment is taken
typically higher than the bending moment corresponding to the onset of first yielding (consequent with equa-
tions (4.4) and (4.7)).

12
 Figure 4.4 Pile cap moment-curvature (From Pianc (2006))

The simplified bi-linear moment-curvature relations for the plastic hinges accordingly are defined as:
Mn Mu − Mn
( EI )i = for ϕ < ϕ y and ( EI ) p = for ϕ y < ϕ < ϕls (4.10)
ϕy ϕls − ϕ y

4.3.3.6 Conclusions on performance objectives

According to the considered seismic design standards, the most important concluding remarks regarding per-
formance objectives for pile-deck port structures are:
- All performance objectives seem to overlap for a quite substantial part and all are defined according to
the generally adopted dual level design approach;
- Pure elastic response is not a limit state to be considered, even for the low level earthquake
Therefore, using elastic stiffness as a starting point in design is not the best choice. Hence, start with
designing for high level earthquake and then checking for low level seems the best approach;
- Limit material strains are provided by most seismic design standards. Starting from these limit strains
and plastic hinge lengths, maximum allowable rotations in structural nodes can be determined. Defin-
ing appropriate plastic hinge lengths may be quite difficult, estimations may be based on the theory of
plasticity, (Hartsuijker, 2004);
- Maximum allowable deck displacements for jetties seem to be governed by the jetty use and installa-
tions on top of it. Limit displacements therefore seem to partly be supplied by the owner of the struc-
ture, based on a risk assessment.
- PIANC guidelines define not only performance objectives during a seismic event, but also limit residual
displacements. Past earthquakes have shown that these residual displacements for port structures are
greatly affected by soil failure and foundations-soil interaction, for which performance objectives are
limited since these topics are less significant for buildings. Simplified techniques do not include such
soil failure related problems. Hence, nowadays finite element dynamic analysis of the entire soil-
structure system by making use of advanced soil models is getting more common. These modern
techniques generally provide better insight on deformations and stresses along the structure and
eventually result more economic designs and a clearer view on possible failure modes, among other
advantages.
- Past earthquakes have resulted many cases of jetty pile failure due to unacceptable bending moments
in the piles induced by combined kinematic and inertial action, resulting unacceptable permanent de-
formations. For typical jetty structure configuration with the deck having a much higher bending stiff-
ness compared to the piles, extreme bending moments in piles will be located first at the pile-deck
connection and subsequently a in-ground plastic will be formed. With respect to reparability, most

13
 seismic design standards do not allow for the in-ground plastic hinge to form, where at the pile-deck
connection a considerable degree of pile plastic deformations is allowed.

Although MOTEMS specifies performance objectives in a quite detailed way, it is relatively limited in describing
how to implement these performance objectives. A valuable complementary document in this sense is the
book Displacement-Based Seismic Design of Structures ((Priestley, Calvi, & Kowalski, 2007). The book is not an
legally established design standard, but does completely comply with the MOTEMS standard. It provides per-
formance objectives identical to the performance objectives as presented in MOTEMS and besides also exten-
sively describes how to practically implement these requirements in seismic design.
Additional recommendations for determination of ductility capacity levels for pile deck systems are provided by
Goel (Goel, 2010). A simplified procedure is proposed in which the embedment of the pile in the soil is simpli-
fied by assuming an equivalent fixity depth. Subsequently ductility capacity expressions are developed based
for different boundary conditions for the pile-deck connection. It should be noted that assumptions underlying
this simplified procedure should be well verified. Additionally one should realize that the procedure is very
new, and additional research and verification against test results is recommended to establish its accuracy.

4.4 Available seismic design methods

Seismic design of port structures is a discipline where geotechnical and structural earthquake engineering are
involved. Consequently for these projects structural and geotechnical design procedures need to be combined.
The two groups of design procedures are considered in this paragraph in succession.

4.4.1 Geotechnical soil site response analysis

Seismic excitation levels at bedrock level are provided from global seismic hazard maps or preferably from a
site specific probabilistic seismic hazard assessment. In this study the common engineering approach consider-
ing only vertically propagating shear waves is followed, where surface waves are not considered. Then surface
motion levels are calculated based on local soil conditions that determine the transfer function for surface mo-
tion due to bedrock motion. Often for computational convenience a 1D problem is considered which then is
considered to behave as an (equivalent) linear Kelvin-Voigt solid, for which the 1D shear wave equation can be
written as:
∂ 2u ∂ 2u ∂ 3u
ρ = G + η (4.11)
∂t 2 ∂z 2 ∂z 2 ∂t
Which has the a general solution of the form:
* *
u ( z , t ) = Aei (ωt + k z ) + Bei (ωt − k z ) (4.12)
*
Where k is a complex wave number and the two terms relate to upward and downward propagating waves
respectively. The constants A and B are determined based on interface and boundary conditions at the soil
layer interfaces and the surface and bedrock boundaries of the soil deposit.

4.4.1.1 Roesset’s site response amplification function approximation

Roesset’s site response amplification function (Roesset, 1970) for motion amplitudes at surface level due to ex-
citation at bedrock level is a solution of the problem of vertically propagating shear waves through a homoge-
neous soil deposit as presented above, and is defined as:
2
A( f ) = (4.13)
(1 + α ) e ikH
+ (1 − α ) e − ikH
ρ soil vsoil (1 + ξ soil )
Where α= is the soil-rock impedance as a function of soil and rock density, shear
ρbedrock vbedrock (1 + ξbedrock )
ω
wave velocity and damping ratio respectively, k = is the wavelength as a function of frequency
vsoil (1 + ξ soil )
and H is the soil deposit height.

14
When one then assumes the bedrock to be rigid, and the damping ratio to be relatively small, the Roesset am-
plification function can be simplified to:
1
A( f ) (4.14)
2
 H  H ⋅ξ 
cos 2  2π f  +  2π f 
 vs   vs 

Where this amplification function is a simplification of the complex Roesset solution of a single soil layer prob-
lem where analogies of complex power functions and goniometric functions are utilized.
If a soil profile is homogeneous or close to homogeneous and ground motion levels are not very high, Roessets
approximation may be considered to provide reasonable estimates of actual site response. However, often
these conditions are not met in practice and the soil layering is preferred to be included in the site response
transfer function.

4.4.1.2 1D equivalent linear multi-layered soil site response analysis

Nonlinear multi-layered analysis conceptually provides the most accurate transfer function for surface motion
due to bedrock motion, but computational effort is often relatively high and validation of results by a more
simple methods is to be preferred in order to eliminate numerical errors.
For this purpose a 1D equivalent linear multi-layered soil site response analysis in the frequency domain is of-
ten used. It is noted that this analysis type does not account for boundary disturbances and 2D or 3D wave
propagation. However, when a continuous soil profile reasonably may be assumed, the 1D solution may be ex-
pected to be close to the finite element analysis results, provided that finite element model boundary effects
are limited. The 1D multilayered linear analysis configuration is depicted in Figure 4.5, where for equivalent lin-
ear analysis soil parameters are iteratively determined based on average strain levels.

Figure 4.5: Multi-layered linear analysis configuration (after Kramer, 1996)

For a multi-layered soil deposit, transfer functions are derived based on boundary conditions at surface level
and interface conditions related to compatibility requirements at the layer interfaces, for which one is referred
to (Kramer, 1996).

The multi-layered linear approach is extended to an equivalent linear approach by including an iterative proce-
dure for convergence of strain level and related shear modulus and critical damping percentage as is observed
for soils. The representative effective strain level according to (Idriss and Sun, 1992) is determined to be ap-
proximately
γ eff ,layer = Rγ γ max,layer (4.15)
Where γ max,layer [-] is the maximum shear strain observed in a soil layer during the considered time history and

Rγ = ( M − 1) /10 a dimensionless reduction factor, with M being the earthquake magnitude measured ac-
cording to the Richter scale.

15
4.4.1.3 Finite element free field site response analysis
2D or 3D finite element models for free field site response analysis account for different types of waves to de-
velop in the continuum, which by definition will give a more accurate representation of actual surface motions
due to bedrock excitation compared to the 1D approximation. Besides, in such analysis 2D/3D site geometric
properties (e.g. slopes, basins, discontinuous soil profiles) affecting the wave field can be included. By using a
nonlinear advanced soil model it additionally is possible to account for a nonlinear soil constitutive behaviour,
which is not included in a simplified (equivalent) linear approximation. Drawbacks of the potentially more accu-
rate nonlinear analysis are the higher accuracy and the difficulties in modelling nonlinear behaviour accurately.

Finite element free field site response analysis is in practice performed to validate the finite element models
with respect to disturbances by model boundaries and sensitivity to mesh coarseness. It is a preliminary step
towards coupled dynamic analysis of site and structure. When a proper dynamic free field model is obtained,
the structure can be included and the response of the coupled soil-structure system can be calculated.

4.4.2 Structural seismic design methods

Seismic standards and guidelines allow structural engineers to use many different structural analysis methods,
with varying complexities. It is often preferable to apply analysis methods with increasing levels of complexity
levels in succession, in order to optimize the design procedure in terms of efficiency and final accuracy. Relative
simple methods are then used in preliminary design stages, followed by more advanced methods for final de-
sign. Generally the complexity of the recommended analysis procedure increases with increasing ground mo-
tion level and increasing structure importance. Roughly the structural analysis methods as included in different
seismic codes can be subdivided in three main categories, according to PIANC:
- Simplified analysis
- Simplified dynamic analysis
- Dynamic analysis
Furthermore, along performance based design principles, design methods can be divided in methods to be
used to determine structural deformation capacity and methods to be used to determine structural deforma-
tion demand.

4.4.2.1 Simplified analysis

Different types of simplified analysis procedures may be distinguished. However almost all are based on elastic
pseudo-static loads, combined with code specified factors accounting for structural nonlinear response.
In resistance based design, simplified analysis the order of magnitude of the elastic response. Subsequently the
response beyond the elastic range is estimated based on the obtained elastic response and code-specified mul-
tiplication factors.
In performance based design, the simplified analysis procedures are used to estimate structural geometry
based on allowed limit performance objectives (e.g. strains, displacements or drifts) by an iterative procedure
optimizing the stiffness/ductility properties of the structure.

Traditional methods estimate pseudo-static loads based on the soil characteristics of the upper 30 meters of
the soil profile, resulting in a response spectrum representing the response of a SDOF oscillator at the soil sur-
face due to an excitation at bedrock level. Most seismic standards provide design response accelerations in
such a spectral format. Based on local soil conditions an elastic design spectrum for a single degree of freedom
system with damping ratio 0.05 is provided, from which the design structural acceleration can be read based
on the structure’s eigen-frequency. Although seismic elastic design spectra seem to differ considerably over dif-
ferent standards, the key features are similar for all of them. Stiff soil conditions transfer energy to the higher
frequency range, where less stiff soils shift spectral energy to the lower frequencies

Design elastic response spectra provided in most seismic standards are based on 5% critical damping. Some
standards specify how to account for other structural %-critical damping values, which can be very beneficial
since higher damping values lower the design response spectrum.
Additionally most seismic design codes allow the design elastic response spectra to be reduced for behaviour
beyond the elastic range. For example Eurocode 8 includes a behaviour factor q to account for this, while ISO

16
standards make use of seismic reserve capacity factor Cr to distinguish nearly elastic response level earth-
quakes from very rare infrequent high-intensity earthquakes.

Figure 4.6 shows the Eurocode 8 response spectra, indicating the spectral dependence on local soil conditions.
The period at the horizontal axis is related to the SDOF natural period, and Sa at the vertical axis is the SDOF
maximum absolute acceleration for the governing ground acceleration amplitude.

Figure 4.6: NEN-EN-1998:2005 Elastic horizontal response spectra

According to PIANC, simplified analysis is of limited value for jetty structure design in high seismicity areas.
Most seismic design standards (PIANC, MOTEMS, Turkish Technical Seismic Regulations) do not recommend or
even not allow these most simplified methods to be used for port structures in high importance/risk classes or
structures in high seismicity areas. Important factors herein are the effects of soil nonlinearity and coupling of
structure-foundation-soil which is of particular importance for pile-deck systems in soft soil conditions.

4.4.2.2 Simplified dynamic analysis

In simplified dynamic analysis the structural behaviour beyond the elastic range is analyzed by means of a
pushover analysis. In this type of analysis the structure is loaded incrementally up to values beyond its elastic
response range. By this means the elastic and post-elastic behaviour is determined, which subsequently is ac-
counted for in the response spectrum procedure. More specifically, the elastic response spectrum is reduced
based on ductile structural behaviour and additionally, secant structural dynamic properties are determined
from the secant stiffness of the structure which provided the secant structure fundamental period. The accu-
racy of these simplified dynamic type of analyses is relatively well established for multi-story buildings, but less
extensive research has been dedicated to its performance for port structures including pile-deck systems.

Various simplified dynamic procedures with different levels of complexity have been proposed by various re-
searchers over the past decades, among the most important contributions are summarized below:

- Capacity spectrum method (CSM) (Freeman, 1998; Freeman, 2000)

This method was the first such type method. Freeman in this sense is considered the pioneer in this field. The
method uses pushover, combined with response spectra including various damping values. The capacity spec-
trum method was the first method to use the acceleration-displacement-response-spectrum (ADRS) format.
However, there is no clear link provided between ductility from pushover to spectra reduction by damping.

- N2 method (Fajfar, 1999)

The N2-method as proposed by Fajfar, is very similar to the CSM method. Fundamental difference is that in the
N2 method, ductility obtained from push-over analysis is linked to spectra reduction by means of the equal dis-
placement rule. Fajfar has included in his paper several limitations of the N2-method, which are partly related
to this equal-displacement rule assumption. Generally is has been proven that the method seems to provide
relatively accurate results in the medium to long period range, but performs significantly worse in the low pe-
riod range. Additionally the method seems to underestimate responses for very soft local soil conditions. The

17
static nonlinear method as proposed by Fajfar and others, relies on the coupling of demand and capacity via
the discussed relation between force reduction by Rμ and displacement increase beyond the elastic range by μ.
MOTEMS chapter 3104F.2.3.2.5 provides additional relationships dealing with the correct estimate of ductility
and spectral damping ratios.

- Multi-mode (adaptive) methods (Chopra & Goel, 2002; Chopra & Goel, 2004; Chopra, Goel, &
Chintanapakdee, 2004; Aydinoglu & Fahjan, 2003; Kalkan & Kunnath, 2006)
The two methods describe previously estimate seismic demand based on the first natural frequency only. Past
research has shown that this single-mode approach may introduce errors when higher modes contribute sig-
nificantly to structural response. The cumulative participating modal mass to be included is set to 90% in most
seismic design standards, requiring so called multi-mode push-over response spectrum methods. Among the
most important contributions is the research done by Chopra and Goel (2002, 2004), who proposed a modal
pushover response spectrum procedure and showed that this would increase accuracy of predicted response
for reinforced concrete multi-story buildings. Later on Kalkan & Kunnath (2006) proposed an adaptive variant
of this modal procedure, taking into account changing structural properties during an seismic event. It is noted
that despite this adaptive character seems to make the method more accurate, it also complicates the nonlin-
ear static method and increases it’s computational demand. This makes it less attractive for practical engineer-
ing purposes.

The nonlinear modelling tools used for pushover analysis in simplified dynamic analyses are most often struc-
tural engineering finite element software packages (e.g. Scia Engineer, USFOS or Seismostruct), that are capa-
ble of accurately modelling all important nonlinearities in the structure. However, for jetty structures it also
seems reasonable to use geotechnical engineering software like Plaxis in order to better model soil-structure
interaction and allow for interaction of site and structure response.

4.4.2.3 Dynamic analysis

The method to be used for dynamic analysis of structures depends on the level of (non)linearity in structural
behaviour. Assuming a structure to behave as a linear system is generally reasonable in the range of relatively
low load levels, where nonlinear behaviour of the system is limited.
For dynamic analysis of linear systems the following two techniques are commonly used:
- Modal analysis, when damping is assumed to have the Rayleigh damping format
- Fourier analysis, when damping does not fit to the Rayleigh damping format.
The assumption of Rayleigh damping will generally not be appropriate if the system to be analysed exist out
more parts with significantly different stiffness and damping characteristics. Coupled soil-structure systems are
typically such systems where for the soil a much higher damping ratio will be realistic than for the structure.

The techniques available for analysis of linear systems are computationally much more convenient compared
to the analysis of nonlinear systems. For this reason it is often attempted to define an equivalent linear system
that approximates the response characteristics of the true nonlinear system. Major drawbacks of the equiva-
lent linear approximation are the difficulties in obtaining accurate equivalent stiffness and damping character-
istics, since these are depending on the varying load level. High intensity earthquakes generally result in struc-
tural responses beyond the elastic range, and consequently the assumption of equivalent linearity has limited
applicability.

In dynamic analysis of nonlinear systems, the structural response is calculated by nonlinear numerical time in-
tegration. A powerful tool available is the finite element formulation of the problem. Accuracy and stability of
the time integration process depend on the formulation of the time integration. Explicit and implicit schemes
are available. For stability reasons generally an implicit scheme is preferred. A commonly used family of implicit
time integration schemes is the Newmark (1959) type family, that is based on the following equations:
 1  
u t +∆t = u t + u& t ∆t +   − α N  u&&t + α N u&&t +∆t  ∆t 2
 2   (4.16)
u& t +∆t
= u& + ( (1 − β N ) u&& + β N u&&
t t t +∆t
) ∆t

18
Where α N and β N are the Newmark coefficients that determine the accuracy and stability of numerical time
integration. Combined with the dynamic equilibrium equation at the end of the considered time step:
m ⋅ u&&t +∆t + c ⋅ u& t +∆t + k ⋅ u = F (4.17)
t +∆t t +∆t t +∆t
These equations provide the basis for computing u && , u& and u from the known acceleration, velocity
and displacement at the end of the previous time step.

Providing the input of the finite element model to be accurate, dynamic time integration analysis conceptually
is the most accurate tool available to analyse the response of a nonlinear system. However, as recognized in
most seismic design standards, the sensitivity of structural response to the input seismic motion is high. Hence,
it is always to be required to perform the analysis for a number of different seismic input signals.

4.4.2.4 Structural analysis method selection

Along the principles of performance based seismic design, the following statements regarding the preferred
design methods to be used can be made.
Regarding the deformation capacity of structures:
- Deformation capacity of structures can best be demonstrated by a static nonlinear pushover analysis.
When advanced models are used, the actual structural response can be very well approximated by
nonlinear pushover analysis.
Regarding the deformation demand of structures:
- The degree and importance of nonlinearity in the system are decisive factors for the type of deforma-
tion demand analysis method to be applied.
- For linear (or almost linear) systems, seismic deformation demand may be calculated by using modal
analysis or Fourier frequency domain analysis of the (equivalent) linear system, for Rayleigh or non-
Rayleigh damping respectively.
- For nonlinear systems of which the response is governed by the fundamental mode, simplified dy-
namic single mode response spectrum methods will provide reasonable results. A related minimum
participating modal mass of the first mode is set to 90% by most seismic standards.
- For nonlinear systems of which higher modes contribute significantly to the response, multi-mode
simplified dynamic response spectrum methods are generally more appropriate compared to a single
mode approximation. However, the modal combination rule to be used is very much structure de-
pendent (Chopra, 2001), which introduces a significant uncertainty in the results
- Using an adaptive variant of the multi-mode simplified dynamic response spectrum methods may in-
crease accuracy of calculated demand. However, it also significantly increases computational demand.
Consequently one may doubt whether this extended simplified dynamic analysis method is preferable
instead of the nonlinear dynamic analysis.
- Nonlinear dynamic analysis is a powerful tool for calculation structural deformation demand. Load de-
formation characteristics can be implemented varying with load level, as is for actual structures. Using
advanced material models allows for inherently including hysteretic damping in the system. Two
drawbacks are the high computational costs and its sensitivity to the selected seismic input signal. The
former drawback is becoming less critical with increasing computational capacity. The latter is over-
come by selecting a number of seismic input signals and use an envelope as seismic design demand.

4.5 Jetty structural dynamics (under earthquake excitation)

4.5.1 Mathematical representation of systems under seismic loading

Dynamic characteristics of a structure are determined by its geometry and the distribution of mass, stiffness
and damping over the structure. In order to analyse the structure, a representative mathematical model needs
to be defined. The structure geometry is discretised and the structure boundaries have to be defined. Mass,
stiffness and damping are assigned to the structure degrees of freedom, which then results the multi-degree
of freedom equation of motion in matrix notation:

[ M ]u&& + [C ]u& + [ K ] u = F (t ) (4.18)

19
Where:
[M ] = The mass matrix

[C ] = The damping matrix

[K ] = The stiffness matrix

u&&, u&, u = The acceleration, velocity and displacement vectors
F (t ) = The external force vector

In case of earthquake loading, the structure response to imposed ground motions is to be computed. Elastic
and damping forces are then related to relative motion levels of the structure to the ground, however internal
forces are linear with the absolute accelerations. From these observations equation of motion (4.18) can be
transformed into equation of motion (4.19):
[ M ] u&&(t ) + [C ] u& (t ) + [ K ] u (t ) = − [ M ]ι u&&g (t ) (4.19)
&&(t ) and
Where u u&&g (t ) are the structure acceleration vector and ground acceleration varying in time, u& (t ) and
u (t ) are the structure relative velocity and displacement vectors and ι is the influence vector having entries
equal to 1 for al degrees of freedom in the direction on ground motion and 0 otherwise.

4.5.2 Coupling of superstructure, foundation and soil

For building structures the traditional approach was to define the system boundary at foundation level, so
foundations and structure-soil interaction are not included in the dynamic mathematical models. Where this
was believed to give proper results, nowadays according to performance based design principles it is believed
that nonlinear soil foundation structure interaction may be very significant, beneficial or detrimental and hence
should be taken into consideration (Gazetas, 2012).

For jetties and other geotechnical structures with heavy deep piled foundations forming a very prominent part
of the entire structure, the need for inclusion of foundation and soil in design is much more obvious. For jetty
structures where piles are monolithically forming both an important part of the superstructure and the founda-
tion it is clear that one needs to consider the pile in jetty dynamic analysis.
Analysing the response of the piles themselves during earthquakes is traditionally done by following a two-way
approach that is based on superposition of kinematic and inertial effects in order to obtain the governing
maximum bending moments in the piles. However, since for jetties the piles are such a prominent structural
part, coupled analysis seems to be preferred for these type of structures, in order to obtain reliable global jetty
responses. The coupled analysis directly will provide pile bending moments that are resulting from combined
kinematic and inertial loading at the piles and hence will be the governing design bending moments resulting
from the seismic load case.

Including the structure-soil interaction effects in the seismic analysis means that the boundaries of the mathe-
matical model will be extended to include soil constitutive behaviour in the analysis. The traditional approach is
to represent the soil by a Winkler bedding supporting the piles. Then along an uncoupled approach, the sup-
port nodes of the Winkler bedding excited by motions based on a free field response analysis. The alternative is
to construct a coupled finite element model both the soil overlying bedrock and the jetty structure, for which
the coupled response to a bedrock excitation is calculated. Both options are schematically shown in Figure 4.7

20
Figure 4.7: Pile as a Beam on Winkler Foundation model approach (left) and Coupled soil deposit + jetty structure finite
element model analysis (right)

4.5.3 Higher mode contributions in response

When a structure is excited by an earthquake motion signal, the total vibration of the structure can be de-
scribed as a superposition of vibrations in the structure its eigen-modes for linear structures. The contribution
of a mode to the total structure response is represented by the modal participation factor or effective modal
mass for a certain direction, corresponding to the mode considered. Modal participation factors are defined as:
φ nT M I
Γn = (4.20)
φ nT M φ n
Where φ n is the mode vector of the nth mode considered, M is the system mass matrix and I is the identity
matrix. Modal participating factors absolute values are dependent of the mode shape normalisation scheme,
for which reason the absolute value of the modal participation factor has not reason. To overcome this prob-
lem effective modal mass may be defined as:
2

mn =
(φ n
T
MI ) (4.21)
T
φn Mφn

The higher this effective modal mass, the more prominent the mode will be present in the total response. The
presence of a specific modal response in the total response is determined by the frequency content of the
ground surface response signal and the sharpness of the input seismic signal. The sharper the input signal is,
the more the response will be like a pure pulse-response, for which higher modes are more prominent in the
response.

Simplified dynamic techniques in earthquake engineering rely on a number of presumed failure modes, which
often is assumed to be represented by the structures’ fundamental mode only. However, restricting the re-
sponse to the fundamental mode only, may introduce quite substantial errors when modal participating
masses of higher modes have substantial values, the input signal is very much pulse-like or a large static com-
ponent is present. Most seismic design standards recommend to take into account sufficient number of modes
up to 90% cumulative participating modal mass.

21
Modern standards restrict the use of these single mode simplified dynamic methods to structures being sym-
metric in their two main horizontal direction, which is related to the recognized severe effects of torsional ef-
fects in structure responses.
This study focuses on soil-pile interaction aspects for jetty structures and rotational effects due to asymmetry
and excitation phase differences for great jetty lengths are not considered. However, the distribution of par-
ticipation modal mass in the transverse jetty direction will be identified for the typical jetty structure with soft
soil conditions, which may be linked to the performance of a fundamental mode simplified dynamic analysis.

Modal analysis for linear systems results a structure dynamic response on its natural modes, where modal am-
plitudes depend on the dynamic excitation. However, the maximum total motion amplitude for transient dy-
namic problems is not determined, since modal responses extremes not necessarily do occur at the same mo-
ment in time. Various approximations are available. Summing all modal maxima gives an upper bound, which is
usually too conservative for design. More accurate estimates of total response based on random vibrations
theory are available. The simplest one sums the squares of the modal maxima and subsequently takes the
square root to determine the extreme total response (SRSS). This procedure seems to give good results if natu-
ral periods are sufficiently separated (by a factor 1.5 according to Kramer, (1996) page 582). As an alternative,
also procedures that account for correlation between modes are available, for cases of closely spaced modes. A
method of this type is the complete quadratic combination (CQC) rule, in combination with the Der Kiureghian
correlation coefficient ((Chopra, 2001) page 557)

It should be noted that both modal combination rules (based on random vibrations theory) and the response
spectra provided in codes, originally related to wide band seismic input signals with relatively long phases of
strong ground motion. Hence, results provided by these methods might be less accurate when these conditions
are not met for a design case. This problem is recognized in various earthquake engineering research papers
and textbooks. Dynamic time history analysis is generally considered to be preferred for these cases (Chopra &
Chintanapakdee, 2001) , see Chopra 2001 page 559)

4.5.4 Damping

4.5.4.1 Actual damping in structures

The structures’ mass and stiffness to be included in the mathematical model (Equation (4.19)) are in general
relatively easy to estimate. In contrast, defining proper damping values is much more difficult. Various fre-
quency dependent and frequency independent sources of damping that are believed to be important for seis-
mic jetty design are:

Frequency dependent:
- Structural material damping
- Damping resulting locally from strongly nonlinear soil behaviour adjacent to a pile
- Soil material damping for saturated permeable soils (from pore fluid moving relative to the soil skele-
ton)
- Soil radiation damping (energy transport through the soil to the far field)
- Hydrodynamic damping (velocity dependent drag forces acting on the pile)
Frequency independent:
- Structural friction damping (frictional damping in e.g. bolted connections)
- Soil material damping for dry and impermeable soils (friction of soil grains moving with respect to each
other)

Including all different damping contributions in a mathematical model for earthquake response analysis should
conceptually result the most accurate estimate of actual response of the structure. However, this approach is
uncommon in engineering practice for the following reasons:
- Correct damping with proper physical background still unknown
- It is a laborious task to properly include all different damping contributions
- Conditioning of damping matrix may become computationally inconvenient
- Common FEA software does not provide the option to define damping such a specific way.

22
Instead, an equivalent Rayleigh critical viscous damping percentage is often defined to approximate the effects
of various damping mechanisms present in the real structure. However, the following related problems can be
defined:
- Actual damping resulting from hysteretic behaviour often is a nonlinear function of motion amplitude,
as a result it is difficult to properly include damping in the Rayleigh format for dynamic problems with
strongly varying motion amplitudes.
- Rayleigh viscous damping by definition is frequency dependent, where actual damping sources not
necessarily are frequency dependent, e.g. hysteresis loops due to nonlinearities is known to be fre-
quency independent.
Computational advantages of using linear equivalent viscous Rayleigh damping traditionally where considered
to outweigh whatever compromises are necessary in the viscous damping approximation. However nowadays
it is believed that in nonlinear dynamic analysis a better approximation of actual damping is obtained when
hysteresis behaviour (by nonlinearities) and some additional Rayleigh damping are combined. The Rayleigh
damping part then is governing the total damping for low level responses, where it becomes negligible com-
pared to the hysteretic energy dissipation for high level responses.

4.5.4.2 Rayleigh damping and Extended Rayleigh damping formulation

The Rayleigh (proportional, classical) damping format, as presented in equation (4.22), is computationally con-
venient as it meets orthogonality properties of the damping matrix.

[C ] = α [ M ] + β [ K ] (4.22)
Where,
[C ] = the material damping matrix for the system

[M ] = the mass matrix for the system

[K ] = the mass matrix for the system

α and β = damping constants
th
As represented in Figure 4.8, the damping ratio for the n mode of the system is:
α β ⋅ ωn
ξn = + (4.23)
2 ⋅ ωn 2
Where,
th
ωn = the n mode natural circular frequency

Figure 4.8: Rayleigh damping over a certain range of frequencies

th
The damping coefficients α and β are usually determined from specific system damping ratio’s of the i and
th
j modes of vibration according to:
2ωiω j
α =ξ
ωi + ω j
(4.24)
2
β=
ωi + ω j

23
Where,
th th
ωi and ω j = the i and j mode natural circular frequency respectively
ξ = the assumed material damping ratio for all the structures’ vibrations modes neces-
sary to ensure reasonable values for all modal damping ratios prominent in structural re-
sponse.

Applying Rayleigh damping will result a diagonal modal damping matrix, so an uncoupled system for which
modal analysis can be applied. The resulting natural modes are similar to the undamped ones. On the contrary,
non-Rayleigh damping results a non-diagonal modal damping matrix, and natural modes may differ from un-
damped ones.

A variant of Rayleigh damping is the stiffness proportional damping, which is obtained by setting Rayleigh pa-
rameters α = 0; β ≠ 0 . This variant is characterized by increasing linear with frequency, and therefore is often
considered a useful measure for damping out high frequency content in numerical response that is not pre-
vented from oscillating by time integration solution procedure.

Since viscous damping conceptually introduces frequency dependent damping in the system, one should be
very careful with defining the Rayleigh damping parameters. A proper damping value over the entire frequency
range of interest should be obtained. An alternatives is available known as Extended Rayleigh Damping or
Caughey damping (Clough & Penzien, 1993), where more than two basis modes/frequencies can be specified
and so the damping matrix is constructed in a different way, aiming at having a more constant damping value
over the frequency range of interest.
N −1 b
[C ] = [ M ] ∑ ab (  M −1  [ K ]) (4.25)
b =0
Where:
N = the number of frequencies/modes incorporated
N −1
1
∑ a ( 2π f )
2b
ab = a scalar facto defined by ξn = b n
4π f n b =0
Although this damping formulation may provide widening of the frequency range with proper damping values,
its drawback is the higher computational demands that will result. (Hashash & Park, 2002)

The Rayleigh damping approach may be very attractive, it physical background is lacking. Furthermore the ef-
fective critical damping percentages are often arbitrarily defined, where instead careful selection is required.
Defining a certain Rayleigh viscous damping factor, results essentially frequency dependent effective damping
over the varying range of frequencies. One should make sure that the effective damping values obtained have
reasonable values, at least within the frequency range of interest.

4.5.4.3 Structural material and friction damping

Structural damping results from micro-scale material straining (material damping) and friction between sur-
faces moving relative with respect to each other (friction damping). The former is related to heat production in
materials due to vibrations which is a phenomena still poorly understood, but clearly a material dependence
can be found. The latter is e.g. provided by work done by friction in e.g. bolted connections or other structural
slipping surfaces.

Generally structural material and friction damping are combined into only one equivalent viscous damping pa-
rameter. The value of the damping parameter may be found based on literature, or be calculated from the
structures global nonlinear pushover curve / global hysteresis loop as proposed by (Priestley et al., 2007) and
(PIANC, 2001) page 229). The area within a hysteresis loop represents the energy dissipated in a cycle by the
viscous damper. Hence it can easily be related to an equivalent viscous damping percentage by relating it to the
maximum potential energy present in the system (the spring) during a cycle.

24
Eurocode 8 recommendations (NEN-EN 1998, 2004) provide an equivalent viscous damping ratio of 5% critical
to be included in seismic analysis, where different values have to be specified. ISO 19902 recommendations
(ISO 19902, 2006) provide maximum modal damping ratios up to 5% for fixed steel offshore structures during
ELE events, where additional damping, e.g. hydrodynamic or soil induced, shall be substantiated by special
studies. For fixed steel offshore structures under ALE events where inelastic behaviour of both structure and
damping is likely, damping values varying up to 10% critical may be applicable. ISO19903 recommendations
(ISO 19903, 2006) provide maximum damping values of 5% critical to simulate damping originating from struc-
tural joint and hydrodynamic energy dissipation for seismic analysis of fixed concrete offshore structures.
Again, higher values according to these regulations shall be justified by special studies

4.5.4.4 Soil radiation damping

Radiation damping is the transport structural vibration energy to the far field, and often referred to in litera-
ture as geometric damping/attenuation. Transport of energy providing the radiation damping effect is only
possible for waves at frequencies beyond the cut-off frequency of the soil which for soil columns can be shown
to be equal to the fundamental frequency of the soil profile (Wolf & Deeks, 2004)

Radiation damping is much stronger for 3D spreading body waves than for 2D spreading surface waves (decay
rates are respectively 1/ r and 1 / r ). As a result surface waves generally are more prominent for structural
response on relative big distances from the fault rupture ((Kramer, 1996), page 179).

In 2D axi-symmetric or 3D geotechnical finite element analysis radiation damping is automatically included,

where in 2D plane strain analysis radiation damping is hardly present. In reality radiation waves may travel to-
wards ‘infinity’ where in geotechnical finite element models model suffer from for computational limitations. In
order to prevent radiating soil waves generated by structure motions to reflect at the boundaries and simulate
radiation damping effects, finite element model boundaries should be set such that radiating waves are
damped by material damping while travelling towards the boundary, or including absorbent boundaries to pre-
vent wave reflection. In finite element models, boundaries cannot be defined too far away for computational
convenience, therefore artificial Rayleigh material damping in combination with absorbing boundaries are of-
ten used to provide the actual attenuation of stress waves to the far field by radiation damping.
In beam on Winkler foundation models, radiation damping may be accounted for by adding a radiation damp-
ing contribution to the total viscous damper parallel to the Winkler spring, as is treated in paragraph 4.6.3.6.

4.5.4.5 Soil material damping

For dry soils soil material damping resulting from heat energy loss due to friction forces between soil grains
when the soil is vibrating. This friction induced type of damping is generally believed to be frequency-
independent (Hardin, 1965), and is characterized by hysteresis soil behaviour. For saturated permeable soils
the damping however is characterized better as viscous frequency dependent damping (Michaels, 1998;
Michaels, 2008) , since it is governed by pore fluid motion through the soil skeleton generating heat energy. For
very large pore spaces the friction between water and soil particles then again drops, decreasing actual fre-
quency dependence of material damping (Stoll, 1985).
Traditionally soil damping was assumed frequency independent because this was convenient in relation to the
Kelvin-Voigt description of soils, as it easily fits into the definition of a constant complex shear modulus. Essen-
tially a equivalent non-viscous viscosity is defined as equivalent viscosity η = 2Gξ / ω resulting a constant
damping ratio. More recently an alternative representation to the Kelvin-Voigt solid (termed the Kelvin-Voigt-
Maxwell-Biot model) was published for the specific case of shear strain (Michaels, 2006a; Michaels, 2006b) For
practical purposes however mostly an equivalent viscous (frequency dependent) damping or a fixed damping
ratio resulting a constant complex shear modulus is defined.

Critical damping percentages can be estimated based on soil shear strain levels, plasticity index and confining
pressure according to (Vucetic & Dobry, 1991; Ishibashi & Zhang, 1993). A similar but more simple hyperbolic
relationship was proposed by proposed by (Hardin & Drnevich, 1972), that also has proven to give accurate es-
timates of real soil behaviour. In Eurocode 8 an approximation of strain dependent soil stiffness and damping
parameters based on ground accelerations as in Table 4.5 is given, from which may be concluded that for high
level earthquakes damping ratios exceeding 0.10 are found.

25
 Table 4.5: Ground acceleration related shear wave velocity, shear modulus and damping characteristics
Ground acceleration ra- damping ratio [-] vs / vs ,max [-] G / G0 [-]
tio ag S / g [-]
0.10 0.03 0.90 ( ± 0.07) 0.80 ( ± 0.10)
0.20 0.06 0.70 ( ± 0.15) 0.50 ( ± 0.20)
0.30 0.10 0.60 ( ± 0.15) 0.36 ( ± 0.20)
Where the ± SD values can be used to introduce a certain level of conservatism, or to account for soil type,
where the higher values are more appropriate for stiffer soil conditions and the lower values are more appro-
priate for softer soil condition, S is the Eurocode 8 soil factor, g the gravity constant, and vs ,max and G0 the
small strain soil parameters.

For soils, damping is considered to be frequency independent, since it is considered to be produced by irre-
versible plastic deformations of the material. Therefore often a equivalent viscosity η = 2Gξ / ω is defined,
which makes the damping in the soil profile frequency independent. The definition of the equivalent viscosity is
advantageous to analyse rate independent damping by frequency domain analysis. However, with transforma-
tion back to the time domain problems arise when using an equivalent viscosity.

4.5.4.6 Damping at the soil-structure interface

Increased soil material damping in the region surrounding the pile is observed. Due to the relatively high de-
formation levels extreme hysteresis behaviour is locally present, resulting from high local soil strain levels and
plastic shearing at the soil-pile interface. Other effects related to e.g. scour for clays may contribute to in-
creased damping values as well.
Within the nonlinear Winkler concept hysteretic loops dissipating energy are included in load displacement
characteristics due to nonlinear stiffness development for cyclic loading-unloading-reloading. The representa-
tive damping value strongly varies with displacement amplitude. In addition a material damping term for the
adjacent soil may be defined, which is often estimated by expressions as are presented in paragraph 4.6.3.6. As
a consequence of highly plastic local soil behaviour adjacent to a pile, radiation damping is reduced by this
zone. Therefore radiation damping is sometimes considered to be present in the far field only. (El Naggar &
Bentley, 2000)

4.5.4.7 Hydrodynamic damping

The unsupported part of jetty piles are surrounded by free water or air. When during an earthquake the rela-
tive velocities of the pile and the water are nonzero, drag and inertia forces will result according to Morrison’s
equation (explained in Dynamics of Offshore Structures (Wilson, 2003), page 25). The drag force varies linearly
with the relative velocity of the pile and the surrounding water. Thereby, it forms an additional viscous damp-
ing. The dependence on relative velocity of the pile and the water makes the amount of damping difficult to
determine during seismic events, since the motion of the water itself is an uncertain input parameter. Hence is
it uncertain whether the hydrodynamic forces continuously damp structure response, or on the contrary at
some moments amplify it.

In this study the focus is on soil structure interaction during earthquake and for simplicity the pile interaction
with water by hydrodynamic excitation or damping are assumed to be zero. Based on literature this appears a
conservative assumption with respect to the total amount of damping included in the model.

4.5.4.8 Numerical damping

Numerical damping has no relation with any physical process in structural/soil dynamics, and may result from a
certain numerical time integration scheme applied in time domain analysis. The time integration scheme ap-
plied affects the stability of results and energy dissipation for certain frequency ranges. Various time integra-
tion schemes are available, a frequently used family of time integration methods is the Newmark family of
which three schemes are presented in table Table 4.6: Time integration schemes. The properties of an integra-
tion method of this family are determined by two parameters, β N and γ N , which affect the stability and the

26
accuracy of the method. Independent of the integration method used, it is in terms of accuracy always wise to
limit dt to 0.1Tn , Tn being the highest mode with significant response contribution. However, for condition-
ally stable methods, stability requirements may require much smaller time stepping, making the method less
convenient for practical purposes. Hence, unconditionally stable methods generally are to be preferred. A gen-
eralization of the Newmark-β methods is the Hilber, Hughes and Taylor (HHT)-α method where parameter α is
introduced and the condition that the acceleration is constant during the time stepping interval is added. The
parameter α modifies the equation of motion as it is solved in the numerical time integration and adds numeri-
nd
cal energy dissipation, however it cannot be predicted as a damping ratio. An unconditionally stable 2 order
2
accurate method is obtained if 0 ≤ α ≤ 0.33; β = (1 + α ) / 4; γ = 1/ 2 + α . For dissipation parameter
α > 0 , frequencies above 1 / 2∆t are damped out with this integration scheme. The performance of the
method appears to be very similar to the use of stiffness proportional damping

Table 4.6: Time integration schemes

Integration scheme
Linear acceleration
method
Constant average accel-
eration method (By de-
fault adopted in Plaxis)
Newmark, HHT- α modi-
fication
Parameters Stability Numerical damping
β N =1/6, γ N =1/2 Conditionally stable No numerical damping
β N =1/4, γ N =1/2 Unconditionally sta- No numerical damping
ble
α =0.1
β N =0.3025, Unconditionally sta- Numerical damping by
ble numerical dissipation pa-
γ N =0.60 rameter α

By default Plaxis numerical time integration adopts the average acceleration method, which does not introduce
numerical damping. No numerical damping has a disadvantage that response contributions of high frequencies
that are resulting from the numerical process are not filtered out. Hence in the present study, seismic bedrock
signals were filtered by a band-pass 0.1 - 15 Hz Butterworth filter. Though, (Sigaran Loria & Jaspers-Focks D.J,
2011) reported in their study the insensitivity of site responses to inclusion of numerical damping. Conse-
quently when numerical instability problems are obtained it seems reasonable to change the numerical inte-
gration scheme to the HHT-α method with numerical dissipation parameter α = 0.1.

4.6 Soil representation in pile-soil interaction modelling

4.6.1 General comments on soil representation in seismic modelling

In this chapter, soil material models are discussed. Material models are described by a set of equations repre-
senting the constitutive behaviour of the material. Strictly, a material model in this sense should describe ma-
terial mechanical properties in terms of stress and strain, as indeed is the case for soil continua models. How-
ever, in relation to laterally loaded piles also a simplified approach is available, in which soil response is de-
scribed in relation to only one kinematic parameter, being the lateral pile deflection.

Response of soil to laterally loaded piles is not only depending on soil mechanical properties, but also by nature
of loading. Where static loading is the simplest and often well be described, long duration cyclic loading gener-
ally may result strong degrading of soil mechanical properties, where for dynamic loading impedance is also
partly determined by a kind of damping part related to loading rate. Seismic loading is somewhat different in
this sense, since it is a short duration, cyclic, high-rate loading. Predicting actual soil response is therefore quite
complicated and therefore often poorly presented in too simple models.

4.6.2 Linearization of soil response

Dynamics of linear systems can be analysed in the frequency domain, which is very beneficial form an compu-
tational point of view. Basically, the constitutive relationship is assumed to be (equivalent) linear corresponding
to a certain load level, where stiffness is adapted to account for nonlinear behaviour and a corresponding
damping percentage is derived, as depicted in Figure 4.9. Equivalent linear techniques have been developed for

27
both soil site response analysis (Schnabel, Seed, & Lysmer, 1972) and pile structure interaction analysis (Novak,
1974; Gazetas & Dobry, 1984a). For both subcategories holds that an iterative procedure is required in order to
include the proper equivalent stiffness and estimated damping in the analysis.

Figure 4.9: Equivalent linear stiffness and damping

4.6.3 The beam on Winkler-foundation concept and nonlinear p-y curves

4.6.3.1 The Winkler concept and p-y curves, an introduction

Interaction of piles and surrounding soil may in its most simple form be represented by the beam on Winkler
foundation model. Fundamental assumption underlying this model is that adjacent soil layers are reacting in-
dependent when subjected to differential loading. Obviously this Winkler hypothesis does not represent real-
ity, but is very convenient from a computational perspective. Probably for this reason it has become very popu-
lar in engineering practice. The nonlinear impedance relation of lateral pile displacement to soil reaction force
in these Winkler models is traditionally done by defining so called p-y curves. The extensive development of
conventional p-y based approaches seems very much supported by limited computational capacity in the past.
Moreover, it is remarkable that the simple static p-y curves are based on only a relatively limited amount of
field and lab experiments, where they are often used beyond the range they were validated for.

Different soil types have different physical properties, resulting different typical soil type related p-y curves.
Expression for p-y curves as a function of representative soil parameters were proposed for some generalized
soil type categories, based on testing projects mostly including full scale tests for some typical pile diameters
and soil conditions. The most widely used p-y curve expressions for soft clay, stiff clay and sand will be pre-
sented in the next sections and used a starting point in the present study.

Ultimate resistance in p-y curves is generally related to the lowest value of two failure mechanisms. A wedge
failure mechanism governing at shallow depths and flow failure around the pile at greater depths. This dual ap-
proach is implemented in most proposed p-y curve expressions.
It is observed that cyclic loading significantly degrades soil mechanical properties, which makes it necessary to
distinguish between monotonic backbone p-y curves and the degraded cyclic p-y curves. As also indicated in
the introduction of this paragraph, cyclic degrading effects are based on long-duration cyclic loading, which
makes it very doubtful if these degradation effects will be present during the typically transient earthquake ex-
citation. Additionally, some researchers proposed to adapt the static p-y expressions for dynamic loading as a
function of loading frequency, as discussed in section 4.6.3.5.

28
4.6.3.2 P-y curves for soft clay
P-y curve expressions for soft clay have been proposed by Matlock (Matlock, 1970) and are adopted in the ISO
recommendations. A summary of the proposed procedure is presented Figure 4.10 and Table 4.7.

The main input soil parameters for soft clay p-y curves are undrained shear strength cu, effective unit weight γ’
and strain parameter ε50, defined as the strain related to half the ultimate soil resistance. Typical ε50 values for
soft clays range about 0,01-0,02 (Matlock, 1970), where Reese and van Ympe (2001) suggests somewhat lower
values. Ultimate resistance is given as a function of depth, and limited by pf = 9*cu to pf = 12*cu, which is still
under discussion. The former expression for pf (Broms, 1964b) is adopted in ISO recommendations, where
more recent research (Randolph & Houlsby, 1984; Murff & Hamilton, 1993; Martin & Randolph, 2006;
Jeanjean, 2009) consider this value to be too conservative.

Cyclic loading of soft clays might result a significant ultimate strength reduction for shallow depths, as recog-
nized by Matlock (1970). The procedure for post-cyclic loading beyond onset of softening is characterized by a
value of p is equal to zero up to the previous maximum deflection, followed by a second branch parallel to the
secant initial stiffness up to intersection with the curve for cyclic loading. More recent studies have shown that
lateral load transfer in clay can degrade significantly more than adopted in ISO guidelines, particularly if the
loading is two-way. At shallow depths remoulding and gapping may occur, where at greater depths overburden
will prevent gapping but remoulding still may cause softening. Some relatively simple mechanical analogues
have been developed (Boulanger, Curras, Kutter, Wilson, & Abghari, 1999), which approximate the complex cy-
clic behaviour of soft soil by a series of nonlinear springs, sliders and dashpots. Matlock (1970) p-y curves have
infinite initial stiffness, which is recognized to result numerical problems in dynamic analysis. This problem may
be solved by the use of an adjusted stiffness ke for the initial portion (e.g. as developed along elastic theory
(Vesic, 1961) or as presented in (DNV OS-J101, 2012) until the point where both curves intersect. The adjust-
ment for the first portion of the p-y curve is included in Figure 4.10: P-y curves for soft clay according to
Matlock (1970)

Table 4.7. The intersection typically corresponds to 0.23 < p/pu < 0.35. Another possible approach is to just ap-
proximate the smooth Matlock p-y curve by a multi-linear fit, having an non-infinite initial stiffness.

Figure 4.10: P-y curves for soft clay according to Matlock (1970)

Table 4.7 P-y curves for soft clay under monotonic (static) and cyclic loading
Monotonic (static) loading
Compute Ultimate soil resistance ( 3 ⋅ cu ,avg ⋅ D + γ '⋅ z ⋅ D + J ⋅ z ⋅ cu ,avg ) ⋅
pu = min  
 9 ⋅ cu ⋅ D 

Determine deflection at one-half the ultimate soil re- y50 = 2.5 ⋅ ε 50 ⋅ D

sistance y50

29
Develop p-y curve expression 1/3
 y 
p = pu ⋅ 0.5 ⋅  
 y50 

Cyclic loading
Develop p-y curves for p<0,72pu as for static loading See table static loading
Calculate transition depth zr 6cu D
zr =
γ ' D + Jcu
Adjustment initial stiffness according to Vesic (1961) 1/12
 E D4  Es
ke = 2 ⋅ 0.65  s  up to intersection
 EI  1 −ν 2
 pile 
with conventional Matlock p-y curve
For depths beyond zr p = 0.72 pu
For depths less than zr p = 0.72 pu at y = 3 y50
z
p = 0.72 pu at y = 15 y50
zr

Where:
cu = Undrained shear strength
D = Pile diameter
J = Soil related constant
(Matlock (1970): 0.5 for soft clay to 0.25 for medium clay)
N = Lateral bearing capacity factor
Broms (1965): N = 9,
Randolph and Housbly (1984), Martin and Randolph (2006), Murff and Hamilton
(1993): 9.14 < N < 11.92, depending on roughness of the pile-soil interface.
pu = Ultimate soil resistance
y50 = Deflection at one-half the ultimate soil resistance
z = Depth
zr = Transition depth (depth where limit value by flow failure mechanism is reached)
γ’ = Effective soil unit weight
ε50 = Strain at one-half the ultimate soil resistance
Matlock (1970): 0.020 for soft remoulded clay, 0.010 for medium clay, 0.005 for
stiff or brittle sensitive clay
Peck et al. 1974 relates ε50 values directly to values for undrained shear strength cu
ke = Initial stiffness adjustment dynamic analysis

4.6.3.3 P-y curves for stiff clay

Common p-y curves for stiff clay are according Reese and Welsh (Reese, 1975). The procedure is summarized in
Table 4.8. It should be noted that these expressions were based on full scale testing of steel pipe piles in stiff
clay which was probably expansive and therefore continued to imbibe water as cycling progressed. Therefore
the use of Reese and Welsh (1975) p-y expressions for cyclic loading may lead to conservative results for many
other clays.

Similar to soft clays, stiff clay ultimate resistance is related to the lowest of wedge failure or flow failure. Stiff
clays typically show a rapid deterioration of ultimate strength when loaded cyclically up to large deflections.
Besides, cyclic loading of stiff clays may induce gapping which in presence of free water may result scouring by
hydraulic action, resulting in further loss of resistance. Cyclic p-y curves for stiff clays without free water being
present generally will not show these typical softening effects, which also was included by Reese and Welsh
(1975). However, this will be of no relevance in this research regarding seismic jetty response since free water
will always be present and therefore will not be discussed in more detail.

30
 Figure 4.11: P-y curves for stiff clay according to Reese and Welsh (1975)

Table 4.8 P-y curves for stiff clay in presence of free water under monotonic (static) and cyclic loading
Monotonic (static) loading
1. Compute ultimate soil resistance 2c D + γ ' Dz + 2.83cu ,avg z 
pu = min  u ,avg 
 11cu D 
2. Establish initial straight line portion p = ks zy for static or
p = kc zy for cyclic
3. Develop p-y curve using the following 1/ 2
 y 
expression p = 0.5 pu   , y50 = ε 50 D
 y50 
4. Develop the second parabolic branch 0,5 1,25
 y   y − As y50 
of the p-y curve (from As*y50 to 6*As*y50) p = 0.5 pu   − 0.055 pu  
 y50   As y50 
5. Develop the straight line branch (from 0.0625
6*As*y50 to 18*As*y50) p = 0.5 pu (6 As )0.5 − 0.411 pu − pu ( y − 6 As y50 )
y50
6. Develop final straight line branch (be- p = 0.5 pu (6 As )0.5 − 0.411 pu − 0.75 pu As
yond 18*As*y50)

Cyclic loading
Execute step 1 to 3 from the static case Follow step 1 to 3 of static curve as presented above
Develop parabolic branch (up to 0,6*yp)   y − 0.45 y 2.5


p = Ac pu 1 −  p
  , y p = 4.1Ac y50
  0.45 y p  
 
Develop straight line branch (from 0,6*yp 0.085
to 1,8*yp) p = 0.936 pu Ac − pu ( y − 0.6 y p )
y50
Develop final straight line branch (beyond 0.102
1,8*As*y50) p = 0.936 pu Ac − pu y p
y50

Where:
As = Constant for static case, as a function of non-dimensional depth
Ac = Constant for cyclic case, as a function of non-dimensional depth
cu = Undrained shear strength

31
ca = Average undrained shear strength over depth z
D = Pile diameter
ks = Initial subgrade reaction constant for static loading
kc = Initial subgrade reaction constant for cyclic loading
(Matlock (1970): 0.5 for soft clay to 0.25 for medium clay)
y50 = Deflection at one-half the ultimate soil resistance
z = Depth
γ’ = Effective soil unit weight
ε50 = Strain at one-half the ultimate soil resistance
(Preferably to be derived from laboratory tests, if not available Reese and van Ympe (2001)
provide values from 0.004 to 0.007 related to average undrained shear strength)

Matlock also proposed p-y curves for stiff clays (Matlock, 1979) by adapting his original p-y curves for soft clays
to account for more brittle behaviour, especially when cyclically loaded. However, his proposal was not tested
against field measurements.

Typically seismic design standards, including ISO recommendations and former API RP2A only give some com-
ments on p-y curves for stiff clay, but do not provide closed form expressions as are presented for soft clay and
sand. This indicates the scattering that is obtained for various stiff clay response characterizations, apparently
making it difficult to establish generalized p-y expressions. .

Care should be taken when dealing with sensitive brittle soft clays, initial stiffness may be relatively high (ε50 ~
0.005) but subsequently an extreme softening behaviour is obtained for these types of clays (Randolph &
Gouvernic, 2011)

It should be noted that the p-y relationships as presented above do relate force per unit length p to deforma-
tion y, by a relation linearly depending on pile diameter D. (O'Neill & Dunnavant, 1984; Dunnavant & O'Neill,
1985; Juirnarongrit & Ashford, 2001) however have shown that this linear relation may not be most appropri-
ate for clays, where y50 was found to decrease for increasing D, resulting a conservative initial stiffness by
adopting conventional p-y expressions for greater pile diameters.

4.6.3.4 P-y curves for sand

The simple plasticity mechanisms that govern the ultimate resistance for clays, cannot be constructed for sand.
Simple lower bounds cannot be constructed because of combined stress field, and upper bound solutions can-
not be defined because normality rule requires dilation to occurs, which is not allowed at great depths where
the constant volume boundary condition applies. As a result it is more difficult to determine the limiting lateral
resistance in sand and ultimate resistance for sand at varying depths is therefore primarily based on empirical
formulations. Lateral bearing factors are then defined as N = Pf / (D * σ’v). Different simple straight forward ex-
pressions for N have been proposed by various researchers over the past, e.g. (Brinch Hansen J., 1961; Broms,
1964a; Barton, 1982; Meyerhof, 1995). (Reese, Cox, & Koop, 1974) did explicitly distinguish between the two
possible failure mechanism mentioned before (i.e. wedge failure or horizontal flow failure) and came up with
some theoretical expressions for ultimate soil resistance extended with empirical adjustment factors. The ex-
pressions for ultimate resistance proposed by Reese et al. (1974). Reese et al. (1974) describe the p-y relation-
ship by three straight line portions and a parabolic curve. A different p-y expression related to pult after Reese
et al (1974), was proposed by O’Neill and Murchison and included in API and ISO recommendations (O'Neill &
Murchison, 1983). The combined procedure (Reese, Cox & Koop. (1974) and O’Neill and Murchison (1983) is
outlined in Table 4.9.

Both Reese et al. (1974) and O’Neill and Murchison (1983) account for deterioration of soil resistance due to
cyclic loading by scaling factors (As, Ac). However, this limited degradation (no softening) of soil properties does
not include possible severe effects of liquefaction in sand, which may additionally be accounted for. Besides,
cyclic loading predominantly in one direction of piles in cohesionless soil typically induces permanent deforma-
tions because soil particles cave in the gapping space at the back of the pile, preventing the pile from moving
back in its initial position. This effect is also not included in the standard p-y curves but may be taken into ac-
count when constructing p-y loops for dynamic analysis (see paragraph 4.6.3.7).

32
 Figure 4.12: Sand p-y curves after Reese et al.(1974) (left) and O’Neill and Murchison (1983) (right)

Table 4.9 P-y curves for sand under monotonic (static) and cyclic loading
Monotonic (static) and (low rate) cyclic loading
1.Theoretical ultimate soil resistance due to wedge pu , s = ( C1 z + C2 D ) γ ' z
failure (pus)
2. Theoretical ultimate soil resistance due to horizon- pu ,d = C3 Dγ ' z
tal flow failure (pud)
3. Governing theoretical ultimate soil resistance (pu) pu = min { pu , s ; pu ,d }
4. Adjustment coefficient for static or cyclic loading  z
As =  3.0 − 0.8  ≥ 0.9
 D
Ac = 0.9
5. Develop p-y curve  kz 
p = Apu tanh  y
 Apu 

Where:
As = Adjustment coefficient for static p-y curves
Ac = Adjustment coefficient for cyclic p-y curves
C1, C2, C3 = Coefficients from figure 17.8.1 ISO19902:2006
k = Initial subgrade reaction constant from table 17.8.3 ISO19902:2006
D = Pile diameter
pus = Theoretical ultimate soil resistance due to wedge failure (Reese et al. (1974))
pud = Theoretical ultimate soil resistance due to horizontal flow failure (Reese et al. (1974))
pu = Theoretical ultimate soil resistance
z = Depth
φ = Soil friction angle
γ’ = Effective soil unit weight

The Reese et al/O’Neill and Murchison p-y curve expressions for sand are adopted in ISO 19902:2006 seismic
design recommendations. However, analysing the lateral bearing capacities for sand conform ISO, one may
conclude that extremely high equivalent Kp-factors may result for greater depths. This is shown in Figure 4.13,
where ISO lateral bearing factors are plotted with lateral bearing factors as proposed by Broms ( N = 3 ⋅ K p )

or Barton ( N = K p2 )

33
 Figure 4.13: Dimensionless ultimate lateral resistance (Pu/D) for piles in sand

From Figure 4.13 one may conclude that for shallow depths up to 7*D, lateral bearing factors conform Broms,
Barton and ISO do match quite well. However at greater depths ISO appears unrealistic, since allowable limit
stress provided exceeds reasonable values based on the Mohr-Coulomb yield criterion by far. It therefore was
decided in the present study to adapt ISO lateral bearing factors for sand layers at greater depths to Barton
bearing factors. On the other hand it may be noted that greater depth soil response is less prominent in pile
head response, especially for relatively small displacements. This implies only a minor influence from the too
stiff adopted deeper sand layers.

Liquefaction is the phenomena of soils (partly) losing their strength temporarily, due to a set up of pore pres-
sure, caused by high rate cyclic loading of the soil. The rise of pore pressure induces a decrease of effective
stress causing the soil to lose its strength and transform into a viscous substance. In general loose and relatively
fine sands are most prone to liquefaction. The simplest approach for lateral pile analysis in the presence of li-
quefiable layers is to neglect contribution of these layers to pile capacity, or even evaluate which additional
loading effects liquefaction of these layers may result. Typically buckling instability of axially loaded piles due to
temporary loss of lateral resistance in liquefied layers may become a problem. Correctly adapting soil response
characteristics for liquefaction effects and implementing these in analyses has proven to be very difficult. Very
recently Bouckovalas has proposed how to possibly implement liquefaction phenomena in the p-y curve format
(Bouckovalas, 2012). These proposed expressions are not evaluated or adopted in this study because they were
proposed when this study was almost finished. However for future projects comparison of responses along
these expressions, advanced soil models allowing for liquefaction and test results of pile in liquefiable soils is
recommended.

4.6.3.5 Dynamic p-y curves

As mentioned before, response to lateral pile movement is depending on the load nature. Generally the re-
sponse for dynamic (high rate) loading is stiffer than the response to a gradually increasing static load. These
effects have been recognized in many studies and resulted in different attempts to develop dynamic stiffness
relations to be used for pseudo-static analysis. El Naggar and Bentley (2000) proposed a closed form empirical
formulation for dynamic single pile p-y curves as a function of a static single pile nonlinear p-y curve, soil type
and loading frequency as:
 2 ωy  
n

pdynamic = pstatic  α + β a0 + κ a0    (4.26)

  D  

Where:
pdynamic = dynamic value of p on the p-y curve at depth z

34
pstatic = static value of p on the p-y curve at depth z according to ISO recommendations
y = pile lateral deflection
a0 = dimensionless frequency of loading, defined as a0 = ω D / (2vs )
ω = circular frequency of loading equal to ω = 2π f
D = pile diameter
α , β ,κ , n = empirical constants related to soil type as presented in Table 4.10Dynamic p-y curve
parameter constants for a range of soil types*, after NCHRP report 461 page 18Table 4.10

pdynamic resulting from equation (4.26) for increasing displacements and high loading frequencies is limited by
the ultimate lateral bearing capacity of the soil.

Table 4.10Dynamic p-y curve parameter constants for a range of soil types*, after NCHRP report 461 page 18
Soil type Typical parameters α β κ n
a0 < a0 >
0.025 0.025
Soft clay cu < 50kPa, vs < 125m / s 1 -180 -200 80 0.18
Medium clay 50 < cu < 100 kPa 1 -120 -360 84 0.19

125 < vs < 175 m / s

Stiff clay cu > 100 kPa, vs > 175 m / s 1 -2900 -828 100 0.19

Medium dense sand 50 < Dr < 85% 1 3320 1640 -100 0.1
(saturated)
125 < vs < 175 m / s
Medium dense sand (un- 50 < Dr < 85% 1 1960 960 -20 0.1
saturated)
125 < vs < 175 m / s
Dense sand (saturated) Dr > 85%, vs > 175 m / s 1 6000 1876 -100 0.15

* D = 0.25, L / D = 40, 0.015 < a0 < 0.225

Note that α is always unity, since it has to be ensured that for ω = 0 the static stiffness is returned from the
expression. El Naggar and Bentley based the empirical parameter constants on tests on a pile with diameter
equal to 0.25 m, and no test results on typical much greater diameter offshore piles are performed. El Naggar
and Bentley considered the stiffness relations for a frequency range 0 – 10 Hz, which they consider to be typical
for earthquakes. They concluded that the proposed dynamic p-y curves resulted responses in good agreement
with predictions based on a two-dimensional analysis. However, it was noted that accuracy is less for very stiff
soil conditions. Accuracy increases for frequencies greater than 4 Hz, since plane strain dynamic stiffness model
assumptions tend to become less accurate when a situation approaches the static limit case. Later on, as re-
ported in NCHRP 2001 report 461, the dynamic p-y curve expressions where verified along test results, again a
good accuracy of the method proposed by El Naggar and Bentley in 2000 for the given range of application was
found.

The dynamic multiplier is a function of frequency, which can easily be established for frequency domain analy-
sis or loading types fixed at a certain frequency. However, since seismic excitation spans a range of frequencies,
it is not suitable for analysis in the time domain. An approximation may be found by establishing the dynamic
multiplier based on the most prominent frequency in the spectrum. A reasonable approximation may be found
in fixing the dynamic stiffness multiplier to the structure fundamental-frequency, the power spectrum mean
frequency defined as:
m1
Ω= (4.27)
m0

35
or to the power spectrum response central frequency, which is defined as:
m2
Ω= (4.28)
m0
Where
ωN

∫ ω G (ω ) dω is the i
i th
mi = order moment of the power response spectrum and
0

ω N = π / ∆t is the Nyquist frequency, being the highest frequency in the Fourier series.

It was investigated how the dynamic multiplication factors according to El Naggar and Bentley develop with
frequency and displacement amplitude. Significant dynamic multipliers where found. As an indication for f
equal to 5 Hz and y ranging from 0.01 to 0.1 m, the multipliers are approximately 2-8, 3-6 and 1-3 for soft clay,
stiff clay and dense sand respectively.

4.6.3.6 Viscous dampers parallel to the p-y springs

In dynamic analysis of seismic pile-soil interaction, damping is included in the soil interaction models. Since
damping tends to increase for increasing loading rates, various researchers (Gazetas & Dobry, 1984b; Nogami,
Otani, Konagai, & Chen, 1992; Makris & Gazetas, 1992; Kavvadas & Gazetas, 1993) proposed to include a vis-
cous damper parallel to the p-y spring element. Others (El Naggar, 1996; El Naggar & Bentley, 2000) have pro-
posed to define separately the near- and far-field contributions for both linear springs and dashpots. Both
methods have resulted relatively accurate predictions of dynamic pile response. However, the amount of
damping in very much simplified models is often quire arbitrarily chosen, because of its dependence on both
displacement amplitude and loading frequency.

Because of the increasing contribution to total stiffness of the frequency dependent imaginary part, El Naggar
and Bentley (see previous paragraph) also attempted to fit expression for dynamic nonlinear stiffness in the
more conventional complex valued stiffness. This results equation (4.29):
pdynamic = (k1 + ik2 ) y = (k + iω c(ω )) y (4.29)

 ωy  
n

pstatic  β a0 2 + κ a0   
  D  
where k =
pstatic
and c(ω ) =

y ωy
The real part k x (ω ) represents dynamic stiffness and the imaginary part cx (ω ) is related to the out-of-phase
frequency dependent damping, which is generally considered a sum of energy dissipation by material hyster-
etic damping in the near field and radiation damping in the far field. However, this approach resulted too high
damping contribution to overall dynamic stiffness for the higher frequencies.

Alternative expressions for dashpot coefficient accounting for the energy dissipation by the soil are proposed
by different researchers. Common practice is to add separate contributions for both hysteretic material damp-
ing and radiation damping as:
cdashpot = cr + cm (4.30)
According to Kagawa and Kraft, the expression for material damping can be related to average shear strain am-
plitude in the soil which they related to local lateral pile displacement by:
 1+υ 
γ avg =   y pile ( z ) (4.31)
 2.5 D 
For which y pile ( z ) is often obtained from simplified dynamic analysis or can be found by a iterative procedure.
Subsequently, a damping ratio corresponding to the average shear strain amplitude may be defined, and the
dashpot coefficient then can be determined as:

36
 ξ
cm = 2ksec ant ( y ) (4.32)
ω
Among the first to present expressions for the radiation damping were Berger et al (Berger, 1977). The Berger
1D model utilizes the analogy with 1D wave radiation in a rod and accounts for radiation of energy in both the
direction of shaking (compression waves) and in the transverse direction (shear waves). Berger proposed as
dashpot parallel to the soil spring (dynamic stiffness), with a dashpot coefficient equal to:
 v 
cr = 2 D ρ vs 1 + p  (4.33)
 vs 
Where v p and vs are related through the soil Poisson’s ratio υ when linear elastic material behaviour is as-
sumed:
2 (1 − υ )
v p = vs (4.34)
1 − 2υ
Consequently, v p tends to infinity if Poisson’s ratio approaches 0.5 (undrained soil material behaviour), which
is not realistic. According to Gazetas and Doby (1984) v p may be better estimated as:
3.4vs
vp = ≈ 2vs (4.35)
π (1 − υ )
Where they used the Lysmer’s analog wave velocity, derived for surface foundations subjected to vertical oscil-
lations. In their study Gazetas and Dobry also proposed an alternative expression, that is based on assuming
radiating waves in four quarter-planes (shear waves for two quarter-planes and compression waves for two
quarter-planes) and assume a horizontal plane-strain situation. Adding up energy that is radiated away in total
will then provide the following expression for the dashpot coefficient:
  3.4 5/ 4  π 3/ 4
 
cr = 2 D ρ s vs 1 +      a0
−1/4
(4.36)
  π (1 − υ )    4 
 
Where a0 = π fD / vs

4.6.3.7 P-y loops for dynamic analysis

The p-y curve expressions as discussed in the previous sections have primarily been developed for static load-
ing, where cyclic variants are just models to be used for again static analysis with reduced stiffness represent-
ing the cyclic nature of loading. These nonlinear curves in fact are hypo-elastic models because unloading oc-
curs along the loading path. When applied in dynamics, these extended conventional static models do not in-
clude hysterese behaviour and possible residual displacements as is observed for real soils and of which the
former is inherently is providing damping.

The damping artificially often is added to the model by introducing equivalent viscous dampers as proposed by
many researchers and discussed before. With respect to peak displacements and peak stresses during seismic
loading this simplifications have proven to result reasonably estimates of actual pile response. However, not al-
lowing for plastic deformations results the ignorance of possible residual displacements, which is problematic
from a performance based design perspective. In order to allow for residual displacements in time domain
analysis using p-y curves, they need to be transformed into p-y loops, properly accounting for hypo-elastic,
plastic and unload/reload behaviour of the soil. Several researchers have attempted to develop proper p-y loop
models by empirical fitting to experiments. A mathematical model allowing for p-y loop definition was devel-
oped (Allotey & El Naggar, 2005) and is schematically presented in Figure 4.14. From the figure it becomes evi-
dent how many input parameters the model is based on, which makes it impractical for design purposes and
very prone to errors.

37
 Figure 4.14: P-y loop mathematical model after Allotey and El Naggar (2005)

Instead, when using pile-soil interaction with advanced well established soil continua models, the loading-
unloading-reloading loops are much more straight forward included based on a lower number and more com-
mon soil input parameters.

4.6.3.8 Additional comments on p-y curves

The p-y curve expressions as defined by ISO recommendations are strictly only valid for homogeneous soil. In
practice, this obviously will not be the case for deep foundations. To deal with this problem, one needs to
adopt the expressions for layering of the soil. This can be done by changing the way in which vertical effective
stress and average cohesion over depth are included in the expressions. In this research the curves adopted for
soil layering will be used for the comparative studies.

The initial stiffness as present in p-y curves always has been under discussion. Various papers and codes recog-
nize this problem and its related sensitivity of results. Additionally, soil degradation and loading history de-
pendence are poorly accounted for in the conventional p-y based approach. It seems always to be advised to
perform a sensitivity analysis of the results obtained in order to ensure proper and safe designs based on the
very much simplified p-y based approach.

4.6.3.9 Pile group effects

Piles installed in groups to form a foundation might interact when loaded, because of overlapping stress field is
the surrounding soil. Interaction effects may reduce lateral stiffness as well lateral ultimate capacity. In general
reductions are related to spacing in the loading direction to pile diameter ratio and pile group geometry. When
analysis piles in groups by means of Winkler models, these efficiency reduction due to piles interacting might
become significant, especially when piles are closely spaced.

Studies have been performed in the past decades, resulting different empirical efficiency factors for leading
and trailing pile rows under lateral loading. Reductions up to 60% for ultimate capacity and 75% for lateral
stiffness and have been suggested by Poulos and Davis (1980) as indicated. NCHRP Report 461 (Transportation
Research Board, 2001) provides even more extreme reductions for subgrade modulus of pile positioned in
groups, decreasing to approximately 0.30 for second or third trailing pile rows. Reese and van Ympe provide
empirical reduction factors for lateral bearing capacity of pile groups as a function of pile spacing (Reese & an
Ympe, 2001).
An Excel sheet was constructed for finding the reduction factors by Reese and van Ympe, related to pile group
geometry. Efficiency reduction factors for two closely positioned piles are shown in Figure 4.15. Based on these
factors the total efficiency for a pile in a group can be found by multiplication of efficiency reduction factors
due to all adjacent piles:

38
 e pile = ∏ eadjacent pile i (4.37)
i =1..n

Figure 4.15: Efficiency reduction factors according to Reese and van Ympe (2001)

4.6.3.10 T-z and Q-z curves for axial pile behaviour

Although this study focuses on lateral behaviour of piles, also the axial response needs to be considered in or-
der to be able to obtain reasonable Winkler foundations for the jetty, where rocking effects introduce axial pile
movements. Axial pile response in the Winkler foundation concept is described by the Qshaft-z and Qtip-z curves.
Qshaft-z curves describe the relation between axial displacement and soil reaction acting at the pile shaft per
unit pile length. Qtip-z curves represent the relation between pile tip displacement and the related soil reaction
force acting at the pile tip. In this study for simplicity linear relationships are assumed for these axial pile im-
pedances, where it is realized that in reality nonlinear characteristics are obtained. A distinction generally is
made between axial response for cohesive and cohesionless soils. For cohesive soils the axial impedance is of-
ten described as a function of stress level and undrained shear strength, where for cohesionless soils axial re-
sponse is found as a function of stress level and empirical bearing factors related to relative density. The linear
impedances adopted in this study are according to ISO-19902 chapter 17.4, to which one is referred for more
details (ISO 19902, 2006).

4.6.4 Continuum material models for soil

4.6.4.1 General comments on continua material models for soil

Mechanical properties of a continua are described by a set of constitutive equations relating stress to strain
tensors, or stress rate tensors to strain rate tensors as is often implemented in numerical procedures. Tradi-
tionally, soil constitutive behaviour was often simplified to just linear elastic perfectly plastic models. Over the
past decades more advanced soil constitutive models have been developed. In this research the Hardening Soil
model (Schanz, Vermeer, & Bonnier, 1999) and Hardening Soil small strain model (Benz, 2006) as implemented
in the finite element code Plaxis are considered. Basic features of these models will be discussed in this para-
graph.

4.6.4.2 Hooke’s law of linear elasticity and the Mohr-Coulomb failure criterion
Hooke’s law of linear elasticity for continua probably is the simplest constitutive model available is to describe
the continuum deformations. However, when deformations increase the assumption of perfect linearity gives a
very poor description of actual soil behaviour. Among the most used theory to account for material nonlinear-

39
ity is the theory of plasticity. Perfectly plastic material behaviour is represented by the Mohr-Coulomb yield cri-
terion, which can be schematically represented in the principal stress space as shown in Figure 4.16. An alter-
native representation of the yield surface for perfect plasticity without the Mohr-Coulomb corner points was
proposed by Drucker and Prager, but is not adopted in this study.

Figure 4.16 The Mohr-Coulomb yield criterion in the 3D principal stress space

Inside the cone representing the yield criterion, material behaviour is essentially elastic. At the cone the mate-
rial behaviour is dominated by plastic straining. Stress configurations outside the cone cannot exist within the
Mohr-Coulomb concept. For more information regarding computational Mohr-Coulomb material modelling
one is referred to (Plaxis Material Models Manual, 2011; Sluys & Borst de, 2010)

4.6.4.3 The Hardening-Soil model

In the Mohr-Coulomb model, material yielding only is a function of the stress tensor. However, for many mate-
rials this actually will not be the case. As a result of plastic straining history, the yield function may develop, in-
stead of being fixed in the principal stress space. This phenomena is called hardening, for which generally two
types are distinguished, being strain hardening and kinematic hardening. Strain hardening is included in the HS
model and can be subdivided in compression hardening and shear hardening. Material models including hard-
ening/softening conceptually better represent actual soil behaviour when implemented in soil finite element
calculations. It is however noted that real soil hardening behaviour is very complex and most models account
for this in a very simplified way. The soil constitutive model including hardening adopted in this study has been
developed (Schanz et al., 1999) and it has been implemented in the Plaxis finite element code. Provided a
proper parameter selection, the HS-model is known to represent soil constitutive behaviour reasonably accu-
rate for both soft and stiffer soil types, which is related to its features of stress dependent stiffness, different
for both virgin loading and unloading/reloading as is also observed for real soils.

In fact the HS model allows for hardening related to material effective stresses, but on top of it a Mohr Cou-
lomb model is applied based on ultimate strength parameters (c’ and ϕ ' ) governing the ultimate failure state
of the material. This is schematically represented in Figure 4.17. Plaxis offers the possibility to input effective
strength/stiffness parameters for both drained and undrained analysis, where undrained material behaviour
then is governed by effective parameters and excess pore pressure generation based on the pore water bulk
modulus (Plaxis Undrained calculation method A). Alternatively one may input effective stiffness parameters
with undrained shear strength (Undrained calculation method B), .which however removes the stress depend-
ent stiffness feature from the constitutive model.

40
Figure 4.17: (left) Expansion of the Mohr-Coulomb yield surface in the HS model for increasing effective stresses
Figure 4.18: (right) The closed HS model yield surface including the yield cap

In the hardening soil model, stiffness develops as a function of the stress/strain state. Three types of soil stiff-
ness (E50, Eoed and Eur) can be inputted independently for three typical loading conditions being triaxial loading,
oedemeter loading and unloading/reloading respectively.

Expansion of the yield surface cone is related to shear strains and stiffness parameter E50 where on the other
hand the expansion of the yield cap (see Figure 4.18) is governed by Eoed and the isotropic stress/strain state.
nc
The shape of the cap closing the elastic region is determined indirectly by K0 and Eoed. Eur governs elastic be-
haviour within the yield contour resulting from previous ultimate stress/strain states.
Stiffness parameters as mentioned before depend on stress conditions, which would imply that a specific stiff-
ness needs to be inputted based on local assumed stress state.

Figure 4.19: Development of Eoed with stress level

Within the Plaxis hardening soil model this cumbersome selection of input parameters is not required. Instead,
the user has to input a reference soil stiffness for a fixed stress level of pref = 100 kPa, and the program auto-
matically adapts stiffness as a function of local stress conditions, the soil ultimate strength properties and a fac-
tor representing the level of stiffness stress dependence.

More details and background regarding the Plaxis Hardening soil model can be found in the Plaxis 2D materials
manual.

4.6.4.4 The hardening soil model with small strain stiffness overlay model
As explained in the previous section, unloading/reloading within the yield surface is assumed to be linear elas-
tic in the original Hardening Soil model. However, truly elastic unloading/reloading behaviour for soils is only
observed in the very small strain range, where stiffness is high. When strains grow, the unloading/reloading
stiffness has a nonlinear dependency on strain amplitude, as indicated in Figure 4.20

41
 Figure 4.20: HSsmall stress dependent stiffness for triaxial loading conditions

A small strain overlay model including this phenomena, has been developed for the HS model (Benz, 2006).
Since the HSsmall model in Plaxis is only an extension of the conventional HS model, it needs only two addi-
tional parameters to be defined, i.e. the very small-strain shear modulus G0 and the shear strain level ϒ0,7 at
which the secant shear modulus Gs is reduced to 72.2% of G0. From various test data, it is found that the stress-
strain curve for small strains can approximately be described by a simple hyperbolic law. The basic characteris-
tic of this hyperbolic relation (Hardin & Drnevich, 1972) is a decrease of stiffness, with increasing strains due to
loss of intermolecular and surface forces within the soil skeleton. As a result, actual soil stiffness also for
unloading/reloading depends on the small strain loading history, which in the HSsmall model is in accordance
to Figure 4.21 taken into account by:
Gs 1
= (4.38)
G0 γ
1+ a
γ 0.7
Where a =0.385, γ 0.7 is the reference strain input parameter as discussed before and γ is a function of the
strain history, determined to be a scalar quantity by Benz (2006). Subsequently the HSsmall model constitutive
soil relation in the small strain range is easily found from the secant shear modulus Gs as determined by equa-
tion (4.38) above, differentiation with respect to strain then results a tangent expression which can be used in
time integration procedures.

Figure 4.21: Small strain stiffness reduction according to the HSsmall model

42
As indicated, only two additional parameters G0 ref and γ 0.7 need to be defined for the HSsmall model com-
pared to the HS model. The former parameter may (just as for the HS model) be inputted related to a reference
stress level and many correlations are presented in literature. A good correlation to void ratio for many soils
was presented (Hardin & Black, 1969) or to unload/reload stiffness as empirically defined by Alpan (Alpan,
1970). The latter reference strain level may be related to the soil plasticity index, or can be defined by using the
original Hardin-Drnevich relationship and the Mohr-Coulomb failure criterion.

Under the dynamic loading, unloading/reloading loops as included in the HSsmall model introduce a hysteretic
damping component. According to the Hardin-Drnevich relationship for G/Go as a function of shear strain, this
damping will be negligibly small for small motion amplitudes, which appears to be unrealistic compared to ac-
tual soil behaviour. Therefore it is recommended (Brinkgreve, Kappert, & Bonnier, 2007) to introduce addition-
ally a small amount of Rayleigh damping in the model. For this Rayleigh damping 1-2% of critical damping may
be assumed to be reasonable. On the contrary, the same study shows the hysteretic damping at higher shear
strain levels resulting from the HSsmall model to be overestimating actual clay material damping, which can be
solved by setting G0 closer to Gur.

More detailed description of the HSsmall model formulation, its background and comments on the input pa-
rameters can be found in the Plaxis Material Models Manual or in the dissertation report by Benz (Benz, 2006).

43
5 Case study project
This Master’s thesis project focuses on a pile soil interaction problem for a case jetty project. In this chapter a
brief description of this case project is presented. For more information one may contact Witteveen+Bos.

5.1 Jetty location and geometry

The new terminal that is intended to be developed is located in the Eastern part of the Sea of Marmara
The jetty will consist of tubular steel piles, supporting a reinforced concrete deck. At the connection pile-deck,
the piles are fixed in the deck which is considered to have a much higher bending stiffness. For the soil-
structure interaction analyses a critical jetty cross-section is selected, as presented in Figure 5.1. The critical
cross-section selected is in deep-water, since displacements are expected to reach highest levels at this loca-
tion and thus will be critical in the displacement based design approach. Extension of the study to shallower
water depths is recommended for future studies.

Figure 5.1: Characteristic jetty cross-section

5.2 Seismic activity at the project location

Seismicity near the intended jetty is governed by a series of tectonically active faults and basins that are lo-
cated at the western end of the North Anatolian Fault. The North Anatolian Fault (NAF) is separating the Anato-
lian micro plate from the Eurasian plate. To the east of the Marmara Sea, the NAF is a single strand with a nar-
row deformation zone. Then NAF splits in western direction into north, middle and south branches distributed
over a 120 km broad zone around the Marmara Sea. This situation is indicated by Figure 5.2. Fault movements
can be characterized as large scale right lateral strike slip movements extending between Eastern Anatolia and
the Aegean Sea. Furthermore, there are some pull-apart depression basins through NAF, one of which is cross-
ing the Izmit Gulf and Marmara Sea.

44
 Figure 5.2: Tectonic plates and faults

- Different seismic hazard studies performed over the past decade, where for this project, according to
Erdik . (Erdik, 2004). the following design bedrock acceleration levels are to be considered:
- ag,bedrock = 0.70 g for a Mean Return Period of 475 years which is equivalent to a probability of ex-
ceedance of 10% in 50 years
- ag,bedrock = 1.02 g for a Mean Return Period of 2475 years which is equivalent to a probability of ex-
ceedance of 2% in 50 years
These numbers show the extremely high seismic activity of the region, requiring the new build structure to
have significant seismic resistance in order to reach acceptable failure probabilities and ensure a proper per-
formance of the structure even for very high earthquake levels.

5.3 Geotechnical characterisation

Geotechnical parameters used in this study are based on soil survey executed. Some of the available soil survey
and laboratory results are doubtful. These difficulties are often obtained for projects, and consequently a pa-
rameter selection sheet (Appendix I) was constructed to enhance proper parameter selection. Several empirical
correlations are included, for which references are presented in the parameter selection sheet as well. Impor-
tant contributions were found from (Lengkeek, 2003) for sand parameters, textbooks of (Schonfield & Wroth,
1968; Kulhawy & Mayne, 1990; Muir Wood, 1990) for clay parameters and (Benz, 2006) regarding small strain
soil behaviour.
For verification of the obtained soil model parameters, the Plaxis Soil Test Module for triaxial conditions was
used. It is noted that verification by the Plaxis Soil Test module not necessarily provides correct soil parameters
for practical application,. Though with inclusion of the correct stress levels, correlations between undrained
and drained stiffness and effective undrained shear strength levels are verified.

The soil characteristic parameters following from the available soil survey and the parameter selection sheet
are summarized in Table 5.1

45
Table 5.1 Soil profile description
Soil profile for GEOSAN Borehole06, UTM35 E: 704396 N:4507802, Seabed level: 18,00 m-MSL
Soil layer C1, clay, soft C2, clay, medium- S3, sand, very C3, clay hard
stiff dense (Engineering bed-
rock level)
Top level layer -18.00 -28.00 -38.00 -48.00
[m+MSL]
3
γeff [kN/m ] 10.9 12.2 18.0
3
γsat [kN/m ] 16.6 17.4 20.2
wn [%] 52.6 43.1 -
PI [%] 45 44 0
φ’ [°] 23 26 35
2
c’ [kN/m ] 2.5 5 0
2
cu [kN/m ] 4 – 21 31 – 50 0 >200
OCR [-] 1.5 1.5 1

Which has resulted the following material input parameters for the Plaxis Hardening Soil materials.

Table 5.2: Plaxis Hardening Soil input parameters

Soil profile for GEOSAN Borehole06, UTM35 E: 704396 N:4507802, Seabed level: 18,00 m-MSL
Soil layer C1, clay, soft C2, clay, medium-stiff S3, sand, very dense
3
γeff [kN/m ] 10.9 12.2 18.0
3
γsat [kN/m ] 16.6 17.4 20.2
φ’ [°] 23 26 35
2
c’ [kN/m ] 2.5 5 0.5
OCR [-] 1.5 1.5 1
K0,oc 0.80 0.75 0.50
m [-] 0.9 0.9 0.5
ref 2
E’50 [kN/m ] 2900 3800 50000
ref 2
Eoed [kN/m ] 1440 1680 58000
ref 2
Eur [kN/m ] 10000 12500 200000
ref 2
Go [MN/m ] 28000 36000 150000
-4
ϒ0.7 [-] 5.3 10 5.0e-4 1.5e-4
ref 2
vs0 [m/s ] 129 142 269
midlayer 2
vs0 [m/s ] 82 144 317
Kh [m/s] 7.5e-10 3.0e-10 1.0e-3
Kv [m/s] 5.0e-10 2.0e-10 1.0e-3

5.4 Seismic site characterization

Ground characterization affects frequency dependent amplification of a site and structures built at this site,
which is represented in a seismic response spectrum. In order to establish the site specific spectral shape, aver-
age soil characteristics over the top 30 m are generally determined. The top 30 m. of the soil profile (-18,0 +
2
MSL to -48,0 + MSL) is characterized by vs,avg30 = 161 m/s , where vs,avg30 is based on the shear modulus. It is
noted that the shear wave velocity presented in Table 5.2 is related to reference small strain stiffness G0,
where lower values will be found for increasing strain levels. Based on the upper bound shear wave velocities
corresponding to shear stiffness at mid-layer stress levels, the estimated upper bound site fundamental fre-
quency would be:

vs ,avg (82 ⋅10 + 144 ⋅10 + 317 ⋅10) / 30

f0 = = = 1.5Hz (5.1)
4H 4 ⋅ 30

46
The ground profile corresponds to the site classes according to seismic design standards as presented in Table
5.3.

Table 5.3 Ground profile characterization according to seismic standards

Standard Site class
Eurocode D
Turkish code, ISO 19901-2 and NEHRP-FEMA450 E
PIANC indicates that various codes provide characteristic
profiles

Earthquake design procedures according to various standards have been compared in the preliminary study.
The way in which design peak ground accelerations and design spectra are computed appear to be somewhat
different in the various standards, however generally all provide a similar response spectrum approach. Re-
sponse spectra according to the Turkish Seismic Regulations, The U.S. Fema450 and the international ISO stan-
dards are identical, provided that the same input parameters are used, where Eurocode 8 gives somewhat dif-
ferent response spectra. Hence, in this section the response spectra provided by Turkish seismic regulations
(representing also U.S and ISO standards) and Eurocode 8 are presented. The input peak ground accelerations
and site specific spectral input parameters for different earthquake levels are based on a site specific probabil-
istic seismic hazard assessment (Erdik, 2004)

5.4.1 Turkish/ISO seismic regulations response spectra

According to the Turkish seismic regulations, a jetty where hazardous materials or chemicals are handled is a
special class structure, resulting in the following performance objectives:
- No damage for E1 level earthquake
- Minor repairable damage for E2 level earthquake (structural response up to elastic-plastic limit)
- No collapse requirement for E3 level earthquake

The response spectra characteristics related to the three earthquake levels are presented in the tables below.
Bedrock acceleration ag and spectral parameters Ss and S1 are after (Erdik, 2004). The spectra is derived based
on ground acceleration values and local soil classification. The response spectra is then constructed by simple
formula for which one is referred to the related seismic design standard.

The response spectra given by the U.S. Fema450 /MOTEMS and the ISO recommendations are identical to the
spectra conform the Turkish Seismic Regulations, provided that identical input parameters are used. According
to ISO global seismic hazard maps, the governing peak ground acceleration for the site considered should be
less, but the PSHA by Erdik (2004) overrules these lower seismic levels.

Table 5.4: Elastic Design Response Spectrum (5% damping) parameters, site class E
Earthquake level E1 (50% probability of ex- E2 (10% probability of ex- E3 (2% probability of
ceedance in 50 years, ceedance in 50 years, exceedance in 50
MRP = 72 years) MRP = 475 years) years, MRP = 2475
years)
Peak ground acceleration 0.30 0.70 1.02
bedrock ag [g]
Spectral value Ss [g] 0.65 1.54 2.32
Spectral value S1 [g] 0.26 0.70 1.14
Fa [g] 1.4 0.9 0.9
Fv [g] 2.96 2.40 2.40
Sms [g] 0.91 1.39 2.09
Sm1 [g] 0.77 1.68 2.74
Sae [g] 0.91 1.40 2.10
To [s] 0.17 0.24 0.26
Ts [s] 0.85 1.21 1.31

47
5.4.2 Eurocode EN-1998 (2005) response spectra
EN-1998 defines seismic intensity based on two separate performance requirements:
- The Damage Limitation Requirement (DLR)
- The No Collapse Requirement (NCR)
These two limit states are related to seismic return periods and exceedance probabilities. Subsequently these
quantities can be scaled according to a structural importance class.
According to EN-1998 a jetty where potentially hazardous materials or chemicals are handled, should be in
structural importance class III or IV:
- Importance class III: high risk to life and large economic and social consequences of
failure
- Importance class IV: exceptional risk to life and extreme economic and social
consequences of failure
This classification appears to be subjective to the engineers interpretation. For this case the jetty is assumed to
be a class IV structure, resulting an importance factor γ1 = 1.4.
Erdik et al. provides the governing peak ground acceleration based on a site specific probabilistic seismic haz-
ard assessment to be agR,NCR = 0.7 g with a probability of exceedance of 10% in 50 years, which is the reference
situation for the EN-1998 “No Collapse Requirement”. This can be transformed into agR,DLR = 0.41 g for the
“Damage Limitation Requirement” Subsequently both acceleration values need to be scale in order to include
the seismic importance class, resulting:
- ag,DLR = 0.57 g, related to exceedance probability of 17.4 % in 50 years, MRP = 261 years
- ag,NCR = 0.98 g, related to exceedance probability of 3.6% in 50 years, MRP = 1303 years

According to the soil investigation as described in paragraph 5.3, the local conditions are comparable to site
class C of D according to EN-1998.

Table 5.5Elastic Design Response Spectrum (5% damping) parameters, site class D
Earthquake level / DLR, spectrum type DLR, spectrum type NCR, spectrum type NCR, spectrum type
spectrum type 1 2 1 2
Peak ground accel- 0.57 0.57 0.98 0.98
eration bedrock ag
[g]
Soil factor S 1.35 1.8 1.35 1.8
Sae [g] 1.92 2.56 3.30 4.41
TB [s] 0.2 0.1 0.2 0.1
TC [s] 0.8 0.3 0.8 0.3
TD [s] 2.0 1.2 2.0 1.2

It is noted that peak response spectral values conform Eurocode 8 reach higher values than the peak spectral
values resulting from Turkish, U.S. or ISO seismic recommendations. For the case project considered Turkish
and ISO seismic recommendations are valid and accordingly the response spectra provided by these codes will
be adopted in design.

5.5 Jetty structural characteristics

This study focuses primarily on the dynamic soil-structure interaction effects typical for jetty structures. The
structural geometry therefore is taken as a starting point, and based on the Witteveen+Bos design report. The
geometrical properties of the deep water tubular steel piles are presented in Table 5.6.

48
Table 5.6 Pile Geometry
Pile geometry Design value
Sea bottom level [m + MSL] -18.00
Pile top level [m + MSL] 3.80
Pile tip level [m + MSL] -39.00
Pile [-] Ø1372/26
4 4 4 10
Pile moment of inertia [mm ] π/64 x (1372 - 1320 ) = 2.49*10

Pile material parameters used in design are shown in Table 5.7. The material properties shown are representa-
tive values as used for seismic design considerations.

Table 5.7 Material parameters

Pile material parameters Design value
Steel quality [-] S275J0H according to EC-10219;2006
3
Steel mass density [kg/m ] 7860
2
Steel representative yield [N/mm ] 275
strength fy,rep
2
Steel ultimate tensile [N/mm ] 500
strength fu
2
Steel Young’s modulus Es [N/mm ] 210,000

Combining pile geometry and material parameters, the jetty masses to be included in dynamic analysis are de-
fined as presented in Table 5.8.

Table 5.8 Jetty masses for dynamic analysis

Pile dynamic characteristics Design value
1
Pile mass [kg/m ] 7860 x π x 1,372 * 0.026 = 880.8
1 2
Added mass water enclosed in the pile [kg/m ] 1000 x 0.25 x π x 1.372 = 1478.4
1
Total mass pile + water [kg/m ] 880.8 + 1478.4 = 2359
1 2
Added mass soil enclosed in the pile [kg/m ] 1670 x 0.25 x π x 1.372 = 2469.0
1
Total mass pile + soil [kg/m ] 880.8 + 2469.0 = 3350
Representative deck mass [kN/6m 791.8
jetty]
2 6
Pile bending stiffness EIpile [kNm ] 5.23*10

At the pile-deck connections, the piles are typically embedded in the concrete deck beams that are integrated
with the deck slab. The effective stiffness of this concrete section (beams + effective part of the deck) can be
assumed to be much more stiff than the single pile bending stiffness. Therefore, the piles can be considered
fixed for rotations by the deck. The connection pile-deck will be a plugged dowel type connection, built up by a
concrete plug fixed in the steel hollow section, connected to the deck concrete by reinforcement dowels. To
enhance ductile capacity of the structure, it is possible to design the connection to provide additional rotational
capacity and a certain amount of hardening beyond the onset of plastic hinging of the pile. Alternatively the ro-
tational capacity may be provided by plastic hinging of the pile section, for which the highest bending moment
will occur at the connection with the deck. Damage will then be concentrated to this region which can be re-
paired relatively well after a strong earthquake.

In the present study, the design of the ductile connection will not be treated in detail. The post-yield hardening
of the pile at the deck connection is determined based on ultimate acceptable strain levels for the dual level
earthquakes provided by design standards. Based on these strain levels the post-yield hardening capacity and
the related ductility are determined according to the theory of elasticity/plasticity for circular hollow sections,
where a simplified bilinear moment-curvature relations is assumed. Results are presented in Table 5.9.

49
Table 5.9: Bilinear approximation rotational fixity stiffness deck-pile
Parameter Unit Value
Initial fixity stiffness
M y ,initial yield = I y f y / ( D / 2 ) [kNm] 9984

M y ,eff = 1.1M y ,initial yield [kNm] 10980

ϕ y ,eff = 1.1ε y D / 2 [rad/m] 0.002

θ y ,eff = 0.5ϕ y ,eff L fixationlength [rad] 0.001

Where the effective fixation length of the pile in the con-
crete is assumed at 1.00 m
7
K initial = M y ,eff / EI pile [kNm/rad] 10
Post-yield hardening fixity stiffness
∆ε p = ε u − ε y ,eff [-] 0.024

∆ϕ p = ∆ε p D / 2 [rad/m] 0.017

∆θ p = ∆ϕ p L p [rad] 0.025

Where L p is the effective plastic hinge length, which can

be estimated as L p ≈ D or L p ≈ x(α − 1) / α
Where x is the distance over which bending moment
decreases linearly from M max to zero ≈ 15m and
α = M u ,eff / M y ,eff ≈ 1.1 , resulting L p ≈ 1.4
4
K post − yield hardening = ( M u − M y ,eff ) / ∆θ p [kNm/rad] 4.3*10

Post-yield hardening ratio r = K post − yield hardening / K initial [-] 0.004

50
6 Analysis method

6.1 Introduction
The goal of this study is to analyze the operation and result of available performance based seismic design ap-
proaches for jetty structures under soft soil conditions, where according to literature soil structure interaction
is recognized as an important factor. Within the performance based design strategy, the acceptable seismic de-
sign procedures for jetty structures are:
- Simplified dynamic analysis by static pushover analysis combined with a response spectrum procedure
- Two step uncoupled dynamic analysis of the site and the structure
- Coupled dynamic analysis of site and structure system
All three design methodologies will be followed for a case jetty project, as a two-way approach also is recom-
mended in most seismic design standards. Figure 6.1 summarizes the possible analysis procedures along per-
formance based seismic design. Simplified dynamic analysis procedures are often the starting point, where dy-
namic analysis is known to may provide more economical designs where they give more insight in possible fail-
ure modes. Dynamic analysis traditionally is performed for first the site and then the structure separately,
where increasing computational capabilities nowadays also allow for coupled response of site and structure.
Both the uncoupled and the coupled approach will be part of the study. The new released Plaxis 3D dynamics
module is a valuable tool for performing coupled analysis of site and structure. Verification of its operation, ca-
pabilities and limitations in dynamic jetty analysis will be part of this study.

Figure 6.1 summarizes the three possible analysis approaches towards the seismic jetty response, and accord-
ingly are included in this study. It is investigated how the different methods account for soil-structure interac-
tion effects in seismic jetty response analysis. Steps that subsequently will be taken within the three perform-
ance based design approaches are described in the next paragraphs.

It is noted that, as in many other studies, full scale test results for direct verification of the results from the dif-
ferent models are not available, where they obviously would provide very valuable verification. This problem as
is often faced in seismic design is partly overcome by proper step-by-step verification in all steps, which is re-
quired in order to in the end have valuable results.

51
Figure 6.1: Analysis procedure flowchart

52
6.2 Pushover analysis

6.2.1 Static pushover analysis on conventional piles supported by Winkler foundation

models
Static pushover analysis forms the basis for simplified dynamic analysis procedures which is often used for
seismic design in engineering practice. In this stage pushover analysis on piles laterally supported by a nonlin-
ear Winkler foundations will be performed. This simplified Winkler models are often used to represent the soil
because of their computational convenience. Nonlinear Winkler spring characteristics corresponding to the lo-
cal soil conditions will be based on common expressions as presented in paragraph 4.6.3.

Code provisions regarding the spring characteristics of the discrete Winkler foundation are initially adopted and
are discretised as a function of local soil parameters, pile diameter and local depth, with a 1 m spacing. In
analyses Matlock p-y curves for soft clay are used for the very soft C1 clay and the soft C2 clay layers, the
O’Neill and Murchison p-y curves for sand are used for the S3 sand layer. The p-y expressions used do not in-
clude cyclic softening, since the cyclic reduction as obtain for long term unloading/reloading is not to be ex-
pected during an earthquake for the undrained conditions of the clay. Degradations is only to be expected in
the very top soil layers, where resistance will contribute only very limited due to the soft soil present. For input
convenience the nonlinear springs are approximated by equivalent tri-linear springs. The p-y curve characteri-
zations of soil impedance varying with depth are presented in appendix III.

First pushover analysis are performed on single pile models, since in the continuum soil modes group effect
need to be excluded in this initial phase. Subsequently pushover analysis on three-pile jetty cross-section mod-
els are performed. For the single pile models a rotational spring on top is included in the models, having an ini-
tial stiffness equal to the pile bending stiffness and a post-yield stiffness approximately equal to the pile post-
yield bending stiffness, resulting in some hardening after yielding in the pile structural model.

Figure 6.2 Single pile supported by nonlinear Winkler foundation model in Seismostruct

53
6.2.2 Static pushover analysis on a single pile embedded by a continuum with the
Hardening Soil constitutive models in Plaxis
In the present study, the correspondence in lateral pile behaviour between piles supported by conventional
Winkler foundation models and piles embedded in a soil continuum (Plaxis Hardening Soil model) is verified.
The adopted procedure towards a proper well-established Plaxis 3D pile-soil model is explained in this para-
graph.

The HS and HSsmall constitutive models as described in paragraph 4.6.4 are applied in the analyses. Soil input
parameters for the hardening soil models where determined, equivalent to the input parameters for the
Winkler model. The Plaxis models are build up gradually, to prevent inclusion of errors in the models and to
verify its capabilities and limitations. Initially only a lateral load at pile head is included in the pushover models
for computational convenience. Finally a vectorial load pattern will be applied to obtain the approximate load
deformation characteristics of the system for seismic loading.

First a wall with free head is modelled with Plaxis 2D and 3D plane strain models. The wall is given a bending
stiffness equal to the bending stiffness of the pile, in fact representing the typical situation of a row of piles
having centre-to-centre distance of 1 pile diameter. Possible pile modelling by a combined plate+solid is inves-
tigated by checking the performance of a similarly defined wall section in Plaxis 3D. The pushover responses
from the Plaxis 3D models have to be similar in order to establish proper operation of the 3D models, provided
that the more simple 2D model was also built correctly.
As a next step the top rotation fixity of the pile is introduced in the model. The fixation in the single pile model
represents the presence of the deck that prevents pile head rotations for the actual jetty cross-section. The top
fixity included in the models will initially be given elastic characteristics with a very high dummy stiffness. Sub-
sequently an artificial elasto-plastic fixation with hardening will be constructed by making use of elasto-plastic
and elastic fixed-end anchors in Plaxis 3D. This work-around was required since elasto-plastic plates are not in-
cluded in Plaxis 3D. Proper operation of the fixity is verified by comparing these models to models including
rigid fixities and by studying the development of bending moments at pile head.

Figure 6.3: Jetty piles and equivalent half-pile-soil models

Based on the wall models one may conclude whether both the composite plate-solid modelling and the inclu-
sion of elasto-plastic rotational fixation at the head are properly operating. Then both can applied in a 3D single
pile model. For computational convenience, only half a pile is modelled. A model width of 3 m was used be-
cause this is equivalent to a side-side centre-to-centre distance of >4*D for the piles, which based on Reese and
Ympe (2001) would imply that no efficiency reduction will be present. Both a circular and a rectangular pile
configuration with equivalent bending stiffness have been modelled for excluding problems related to the cir-
cular shape modelling in Plaxis 3d. The results confirm proper behaviour of the circular section.

With the pile and fixity properly included in the Plaxis model, soil parameters can be varied in order to establish
the sensitivity of the global structure response to variations in these parameters. Results will be compared to

54
the results from pushover analysis on equivalent pile with Winkler foundation models. Based on this compara-
tive study the equivalence of the models for static loading can be studied, which is a precondition in order to
compare dynamic responses of both models under seismic loading.

6.2.3 Equivalent Plaxis 2d pile model based on the Plaxis 3d and Winkler model push-
over characteristics
Geometric 3D effects important for piles cannot be included in a 2D equivalent model. However, for computa-
tional convenience it would be attractive to define a plane strain Plaxis 2D (wall) model, equivalent to the Plaxis
3D (pile) model. This equivalent Plaxis 2D model then can be used to perform computationally less demanding
2D dynamic analyses instead of 3D dynamic analyses. Additionally it is noted that the Plaxis 2D modelling op-
tions are more complete (e.g. implementation of elasto-plastic structures). This forms a different reason for
which development of equivalent Plaxis 2D pile-soil interaction may be attractive.

In actual pile-soil interaction arching effect in the soil result spreading of the loading over an effective width of
about 2*D to 4*D, depending on the soil internal friction angle ϕ . Arching of stresses in the soil is related to
the concept op group efficiency for pile groups, meaning the reduction of pile capacity due to intersecting load
spreading fields of adjacent piles. The phenomena of group efficiency has been studied by many researchers in
the past. Reese and Van Ympe (2001) have collected these test results and developed efficiency reduction for-
mulas based on these data. In this study a pile efficiency estimate calculation sheet was built, for which one is
referred to appendix III. According to Reese and van Ympe (2001), efficiency of piles will not be reduced for
side-side centre-to-centre distances larger than 4*D, where D is the pile diameter. In order to study the arch-
ing/spreading effects for the soft soil conditions in Plaxis, Plaxis 3D analyses with models of different width
have been performed, implying a pile having a particular side-side spacing from adjacent piles.

Based on the static pushover characteristics compared for both Plaxis 2D and 3D models, it will be judged
whether it is reasonably possible to define an equivalent jetty model in Plaxis 2D.

6.2.4 Pushover analysis of the case study jetty cross-section in transverse direction
In the preceding, the focus has been on lateral response of single piles, and the effects of side-side pile spacing.
For many civil engineering applications, including jetty structures, piles are also spaced in the direction of load-
ing and consequently piles affect each other in this direction too. The side-side pile spacing considered in the
pushover analyses for the jetty cross-section is the actual pile spacing of the case study jetty piles, which is
equal to 6 m. In the transverse jetty direction, which is generally assumed to be the governing direction for
seismic analysis, the pile spacing is 7 m, where the case jetty is supported on three rows of piles. Figure 6.4
shows the 3D and 2D Plaxis jetty cross-section models on the left and right respectively.

Figure 6.4: 3D and 2D Plaxis jetty cross-section models

In Plaxis 3D modelling of elastic-plastic plates is not implemented. Consequently for the jetty cross-section
analysis an artificial elastic-plastic rotation fixity was modelled.

55
Jetty cross-section pushover analysis will provided information regarding pile group efficiency of piles spaced in
the loading direction. Based on Reese and van Ympe (2001) one may expect an efficiency reduction up to 20%
for the trailing piles for the considered jetty design geometry, where other literature (see 4.6.3.9) provide even
higher reductions. Based on the Plaxis 3D analysis these efficiency reductions will be re-assessed for the case
study soft soil conditions.

Subsequently pushover analysis of an equivalent 2D jetty cross-section model will be performed, aiming at cor-
rectly capturing the jetty load deformation characteristics with a 2D model approximation for performing the
dynamic analyses.

Deviations for the 2D model approximations are expected to be caused by two conceptual errors in the 2D
model:
- Load spreading around leading piles is prevented in a 2D approximation, which probably will prevent
trailing piles from mobilizing lateral resistance, ultimately reducing total lateral stiffness for the jetty;
- Local failure effects may be expected to be less pronounced in the 2D approximation, which probably
will increase lateral stiffness.

The simplified Winkler model will result a response identical to the response obtained for a single pile, since
soil response acting the jetty piles is completely independent if p-y element support nodes are not connected
(Figure 6.5). Consequently, the response obtained from a multi-pile jetty cross-section may be expected more
stiff than the equivalent Plaxis 2D and 3D jetty cross-section pushover models. In order to account for group ef-
ficiency reduction of adjacent piles, the required reduction of Winkler soil spring stiffness and ultimate capacity
will be determined based on comparison to the Plaxis 3D jetty pushover analysis results.

Figure 6.5: Jetty cross-section supported by Winkler foundation

6.3 Simplified dynamic analysis based on the push-over results

Recommendations according to the various seismic design standards as presented in chapter 3 of this study will
be followed. The simplified dynamic analysis N2 response spectrum method as proposed by Fajfar (Fajfar,
1999) and summarized in paragraph 4.4.2.2 will be applied. This method is similar to the performance based
simplified dynamic analysis procedure provided by Turkish Technical Seismic regulations for Port and Harbour
structures, appendix C.

The response spectrum according to Turkish Seismic Regulations and ISO (see 5.4.1) is adopted and the govern-
ing peak ground acceleration is found from equivalent linear frequency domain free field site response analysis.
The N2 method as proposed by Fajfar relies on the equal displacement assumption forming the link between
damping and ductility, for which is noted that its accuracy may be less for structures on soft soil conditions. As
an alternative the equal potential energy norm may be adopted, but for the structure fundamental frequency
range typical in jetty design the ductility reduction of response spectra according to either the equal-

56
displacement or the equal-potential energy norm will be reasonably similar according to Miranda & Bertero.
(Miranda & Bertero, 1994)

The number of modes to be included can be estimated by approximating the modal participation factors based
on an equivalent linear system. In this study no multi-mode simplified dynamic analysis will be performed be-
cause of time limitations. The possible improvement in accuracy of the obtained jetty response by applying
multi-mode techniques is recommended for future studies.

Pushover forces are distributed over the jetty transverse cross-section according to the product of the mass
vector and the fundamental eigenvector, where forces related to pile and soil inertia forces below seabed level
will be neglected. These approximations result the commonly used method for calculation of the seismic re-
sponse of multi-story buildings. It is noted that soil-structure interaction by kinematic pile loading is neglected
by this simplified approach. The fundamental mode shape of the jetty structure in transverse direction is de-
termined by modal analysis in Seismostruct, with the soil stiffness being approximated by the initial p-y stiff-
ness.

Better accounting for soil structure interaction in simplified dynamic analysis for the deep founded jetty struc-
ture would require determination of the coupled soil structure dynamic characteristics. In order to do so, a dy-
namic analysis of the coupled system is required, from which the eigenmodes of the coupled system can be
found. Then according to the lowest eigenmodes, the simplified dynamic analysis may be performed. It is noted
that the computational benefits of the simplified dynamic analysis method are then not longer present and
hence this path is not further investigated in this study.

The displacements obtained from the simplified dynamic analysis will be compared to displacements resulting
from uncoupled or coupled dynamic site and structure analyses, from which accuracy of single-mode simplified
dynamic analysis for jetty structures in soft soil conditions can be estimated.

6.4 Free field site response analysis

The free field response for excitation at bedrock level is calculated along the 1D linear elastic frequency domain
analysis and by finite element analysis in Plaxis 2D and 3D. In the Plaxis finite element analysis Linear Elastic
(LE), Mohr-Coulomb (MC), HS and HSsmall material models are used in order to be able to first make a proper
comparison to the frequency domain solution with respect to numerical modelling issues and then study the
effects of the more advance soil models.

A continuous layered soil profile as is assumed to be representative in this study, reduces the free field analysis
to a 1D problem. However, when performing Plaxis finite element analysis, boundary influences are known to
disturb free field site response results within a region around the boundary. In order to exclude these effects to
affect the site response, boundaries have to be taken far enough away and some damping has to be included in
the model preventing waves reflecting at the boundaries to significantly affect the motions at the model cen-
tre. It is noted that a new type free dynamic boundary is available in Plaxis 2D, which conceptually would per-
fectly describe a 1D case. Since these new boundaries are not implemented in Plaxis 3D, a proper comparison
of 2D and 3D models cannot be made when different boundaries are included. Hence in this study the Plaxis
viscous dynamic boundaries are used in both Plaxis 2D and 3D. In a sensitivity study regarding model width, lin-
ear elastic material behaviour will be assumed in order to made a proper comparison with the 1D analytical so-
lution and speed up analyses. A Rayleigh viscous damping ratio of 0.05 is included for the linear elastic soil,
which is believed to be a conservative lower bound value for actual soil damping during moderate and high in-
tensity earthquakes. Sensitivity of model width is studied for both harmonic bedrock signals (3 and 10 Hz) and
the selected Duzce signal. The required model width will be based on analysis of both horizontal and vertical
motions of the surface level from the model centre towards the lateral boundaries.

Both Plaxis 2D and Plaxis 3D finite element analysis are performed in order to assess the performance of the
very new Plaxis 3D dynamic module. Responses for LE, MC, HS and HSsmall soil material models are compared
to find response dependence on soil model type and possible equivalent linear elastic approximations with re-
lated viscous damping percentages.

57
6.4.1 Input signals
In the present study it is assumed that the input bedrock motions are horizontally polarized shear waves
propagating vertically. As input motions both harmonic motions and random earthquake bedrock signals are
used. Only a horizontal bedrock motion in transverse jetty direction is assumed and consequently the Plaxis 3D
free field site response results should match the free field site response as obtained from Plaxis 2D. As a simpli-
fication, the vertical accelerations at bedrock level are in this study set to zero. Additionally Rayleigh surface
waves are not considered too. It is realized that these traditional assumptions in seismic design may be an
over-simplification for a near fault location, but including them is beyond the scope of this study.

The seismic input signals that are selected meet the following selection criteria:
- Preferably the signals should be recorded from the active fault governing the case project location or a
similar type of strike-slip fault;
- Signals should be measured for stiff soil conditions in order to be applicable as a bedrock level excita-
tion;
- Since the case project location can be considered near-fault, a near fault signal should be selected
For the case jetty project two signals from the North Anatolian Fault earthquakes (1999) are selected from the
PEER strong motion database:
- DUZCE_375_N (P1551)
- KOCAELI_IZT090 (P1103)
- KOCAELI_SKR090 (P1109)
These bedrock input signals are presented in spectral format by Figure 6.6. and for representation in the time
domain one is referred to appendix IV.

Figure 6.6: Selected North Anatolian Fault bedrock acceleration spectra Duzce, Kocaeli_IZT and Kocaeli_SKR from PEER
database
As can be obtained from the spectral representations of the signals, the energy is concentrated in the band
ranging from 0.1 to 15 Hz, which is of importance for the finite element mesh size. Based on the frequency do-
main representation of the signals (Figure 6.6) it can be concluded that the Duzce (narrow band) and Kocaeli2
(broad band) signals have very different characteristics, where Kocaeli1 can be characterized somewhere in be-
tween. Consequently dynamic analysis will for time limitations be performed for the Duzce and Kocaeli2 signal
only, where it is noted that in actual design a higher minimum number of seismic input signals is required.

The harmonic acceleration time histories included have a frequency of 3 Hz and 10 Hz and an amplitude of 2
2
m/s , which is a reasonable mean amplitude for the high intensity part of the earthquake motions scaled at
ag,bedrock for the moderate level earthquake according to paragraph 5.2. Harmonic signals are included because
of their simplicity for checking the model performance with respect to mesh size and boundary effects and
make an easy comparison to the analytical solution. A 3 Hz frequency signal is selected because it is a fre-
quency very prominent in the selected real seismic signals. A 10 Hz frequency signal is chosen because it is an

58
upper limit high energy containing frequency of the real seismic spectra, and analysing the response of this
higher harmonic signal can be used to determine allowable mesh element sizes.

6.4.2 Dynamic soil material parameters

As a part of the site response analysis study, a comparative study on results by linear elastic, Mohr-Coulomb
and HSsmall material models is executed in order to analyze the performance of the different models. Actual
soil behaviour in dynamic Plaxis analysis is conceptually best described by using the Hardening Soil Small Strain
stiffness soil constitutive model, provided that soil parameters are properly defined. The HSsmall model in-
cludes stress dependent stiffness development and small strain stiffness features, which introduces hysteretic
damping in the model. However, for verification of the HSsmall model operation, linear elastic and Mohr-
Coulomb soil models are included in this study.

The HS and HSsmall soil models as implemented in Plaxis have a stress dependent stiffness, and consequently
soil layers will not be homogeneous. For linear elastic and Mohr Coulomb models stiffness is assumed constant
over the a soil layer, and fixed at values corresponding to the stress level present at mid-layer depth.

For over-consolidated material subjected to seismic loading the actual shear modulus will develop ranging from
upper limit G0 for small strain levels to lower limit Gur for large strain level, where damping levels will develop
inversely. The resulting lower and upper limit dynamic soil parameters are presented in Table 6.1. For other soil
parameters, one is referred to paragraph 5.3.

Table 6.1: Dynamic soil parameters for LE/MC and HSsmall material models
Parameter Clay C1 Clay C2 Sand S3
3
ρ sat [kg/m ] 1692 1777 2059
Lower bound soil dynamic
characteristics for the
unload/reload range
Emid layer [kPa] 4000 12000 275000

Gur ,mid layer [kPa] 1670 5000 114000

2
vs ,ur [m/s ] 31 53 236
Upper bound soil dynamic
characteristics for the
unload/reload range
G0,ref [kPa] 28000 36000 150000

G0,mid layer [kPa] 11367 36880 207304

γ 0.7 [-] 0.00053 0.00050 0.00015

2
vs ,o [m/s ] 82 144 317

Strain dependent characteristics of soils vary with soil type. Many different curves for Gs / G0 and ξ %crit
based on cyclic strain levels have been proposed. In this study the (Hardin & Drnevich, 1972) relationship (ex-
tended by (Santos & Correia, 2001) and (Benz, 2006)) as adapted in the Plaxis Hardening Soil small strain stiff-
ness model is adopted. Additionally a comparison is made to Ishibashi and Zhang (Ishibashi & Zhang, 1993) who
proposed expressions for Gs / G0 and ξ as a function of plasticity index, confining pressure and shear strain.
These curves have been adopted in the present study and are shown in the next figures for the three present
soils.

59
 Figure 6.7: Shear strain related shear modulus reduction and damping characteristics
note that cut-off Gt,small strain > Gur in the HSsmall model is not included in the graphs

The graphs show the good correspondence of the different modulus reduction curves. The HSsmall Hardin &
Drnevich damping curves for clay seem to somewhat overestimate damping for high strain levels, as was also
noted by Brinkgreve, Kappert and Bonnier (2007) (Brinkgreve et al., 2007). For qualification of the related error
however actual strain levels are required, which will be point of attention in a later stage. The damping levels
for the dense sand after Ishibashi and Zhang turn out to be much higher compared to the HSsmall model in-
cluded damping. However, the very high damping levels based on PI=0 are not believed to be very accurate for
dense sands (oral communication H.J. Lengkeek and D.J. Jaspers-Focks (2012)) and therefore will not be
adopted.

60
6.4.3 Free field site response analysis by 1D Fourier frequency domain analysis
Fourier frequency domain analysis techniques for 1D site response are used in this study as a reference for
numerical analysis. The conceptually linear frequency domain analysis for a multi-layered soil profile is en-
hanced with an iterative procedure to account for strain dependent stiffness and damping characteristics of the
soil. Accuracy of this approximation is verified and conclusions are drawn regarding its usability for predicting
site responses. As a reference for checking operation of (equivalent) linear multilayered response analysis, the
Roesset approximation of the site amplification function (see paragraph 4.4.1.1) was adopted.

Equivalent linear site response analysis of a multi-layered soil profile is in this study coded in Matlab. The Mat-
lab code for 1D analysis of vertically propagating shear waves through a multi-layered soil profile is based on
theory as outlined in the book of Kramer (Kramer, 1996) and briefly summarized in paragraph 4.4.1.2. The Mat-
lab script can be found in appendix VI

It is noted that the frequency domain analysis uses the Fast Fourier Transform algorithm, which is known to in-
troduce errors in violating the response initial conditions because it is based on periodic responses. This effect
is more pronounced for relatively short transient problems with low frequency content and short period of rest
included at the end of the input signal (Chopra, 2001). The previous response period may then not be suffi-
ciently damped out at the start of the considered response. If introduced errors are too significant, the input
signal may be elongated or an improved Fourier method with corrective solution may be applied.

6.4.4 Free field site response analysis by Plaxis 2D/3D finite element analysis
Free field site response finite element analysis is in the present study performed in Plaxis 2D and 3D dynamic
modules. The focus during the free field response analysis in Plaxis 2D is to determine the influence of bound-
ary disturbances, the effects of model width, mesh size and the performance of Plaxis hardening soil models on
site response. In this stage the model will be optimized with respect to these parameters. Subsequently equiva-
lent Plaxis 3D free field analysis is performed which should conceptually give identical results. In order to have
equivalent models in the 2D and 3D modules the following starting points are used:
- Standard viscous absorbent boundaries, and not the improved tied types of absorbent boundaries are
used as a starting point since these are only available in Plaxis 2D and not in Plaxis 3D
- Absorbent boundaries are only applied at the x-min and x-max lateral boundaries. Along the y-min and
y-max lateral boundaries in Plaxis 3D a plane strain boundary condition is assumed as it is in Plaxis 2D
- In both Plaxis 2D and 3D bedrock motions are only applied in x-direction. For the Plaxis 3D models the
bedrock degrees of freedom in y-direction are fixed.
Since the Plaxis 3D dynamics module is very new, the quality of results from the 3D site response analysis is not
very well established an in this study judged by comparison to the 2D equivalent model and the equivalent lin-
ear analytical solution. If based on this comparison the 3D model is believed to give proper results, it can be
used to perform dynamic analysis of the coupled soil-structure system.

6.4.4.1 Mesh element size and critical time stepping

In order to allow for proper wave propagation in the FE model, the maximum element size is restricted accord-
ing to Lysmer and Kuhlmeyer (Lysmer & Kuhlmeyer R.L., 1969) :
λlayer vs ,layer
element sizelayer ≤ = (6.1)
5 5 f max
The stiffness dependency on stress level in the Hardening Soil material models results an actual very low stiff-
ness level in the soil just below seabed level. As a consequence very small elements would be required in this
region, which is computationally inconvenient. The maximum element sizes per layer are in this study deter-
mined based on the mid-layer stress levels as shown in Table 6.2.

The element size based on the very low shear wave velocity for clay C1 is very small and therefore will result
extremely long required calculation time. Therefore it was investigated what would be the effects of a some-
what coarser mesh. The results are shown in paragraph 7.2.1.1.

61
Then, the maximum allowable time step is determined based on the well known Courant’s condition, which re-
stricts the time step by allowing a wave to not travel over more than one element length within a dynamic time
step. The maximum time step herein is based on the compression wave velocity because wave reflections may
result compression wave disturbances in the model that may not give spurious responses.

vc ,layer ∆t element sizelayer

≤ 1 → ∆tmax ≤ (6.2)
element sizelayer 2(1 −ν ur )
vs ,layer
(1 − 2 ⋅ν ur )

Table 6.2: Finite element sizes and dynamic time steps

Parameter Clay C1 Clay C2 Sand S3
Lower limit parameters for LE
and MC soil models
2
vs ,ur [m/s ] 31 53 236

element sizelayer [m] 0.41 0.71 3.14

Upper limit parameters
HSsmall soil model
2
vs ,o [m/s ] 82 144 317

element sizelayer [m] 1.09 1.92 4.22

element sizelayer ,applied [m] 1.00 1.30 1.80

∆tCourant , LE 0.016 0.012 0.004

∆tCourant , HSsmall 0.006 0.004 0.003

Compared to the rule of thumb that dt < 0.1Thighest mod e = 0.1 / f highest mod e = 1/15 = 0.066 , the dynamic
time steps in Table 6.2 can be considered sufficiently small.

It is noted that the default number of required substeps provided by Plaxis is much higher when undrained
analysis is applied for soil. This is explained by Plaxis calculating the number of steps required based on com-
pression wave velocities that tend to infinity when Poisson’s ratio approaches 0.5 as is included in Plaxis for
undrained conditions.

6.4.4.2 Finite element model width

The required model width is determined based on free field analysis for both the two selected Harmonic signals
and the Duzce natural seismic bedrock signal. Criteria used for determination of required model width are:
- Convergence of horizontal accelerations at the centre of the model with the 1D frequency domain so-
lution
- Limited amplitude of vertical oscillations at seabed level in the centre area of the finite element
model, which are resulting from wave reflections at boundaries.

In Plaxis analyses for validation of the required model width, linear elastic material will be adopted for compu-
tational convenience and because plasticity effects will only further limit boundary effects. An linear elastic
stiffness will be estimated based on equivalent linear frequency domain analysis. 5% critical Rayleigh viscous
damping will be adopted. This will give a conservative approximation of the required model width, since actual
soil damping during moderate to high intensity earthquakes will generally be higher, at least for soft soil condi-
tions.

62
6.5 Dynamic analysis of jetty on nonlinear Winkler foundation for support node mo-
tions resulting from free field analysis
In this stage the “structural engineering” approach for dynamic analysis of the jetty structure vibrations in
transverse direction is followed. The response of the structure to motions applied to the Winkler support nodes
will be calculated with finite element code Seismostruct, as indicated in Figure 6.8.

The motions applied to the Winkler support nodes are inputted as acceleration time histories, resulting from
equivalent linear site response analysis in the frequency domain, which results the essentially uncoupled re-
sponse analysis of site and structure. It is chosen to use site response results from frequency domain analysis
because this approach is attractive for its low computational effort and reasonable results for a cases where a
1D approximation of the site is acceptable. By this means a method completely parallel to the computationally
demanding Plaxis finite element analysis will be developed. Additionally, to be able to make proper compari-
son, also the response of the Winkler jetty model for ground motions obtained from Hardening soil finite ele-
ment site analysis was performed. The performance of both methodologies will be resulting from this study.

The Winkler discrete p-y springs applied in the model are identical to the static p-y springs (see appendix III),
where additional dynamic “stiffness” is provided by viscous dampers added parallel to the p-y springs (see ap-
pendix III). Jetty rocking effects introduce axial pile loads although loading will essentially be horizontal. Accord-
ingly axial pile impedances have been defined, where they are for simplicity linearized at the initial stiffness ac-
cording to ISO 19902 chapter 17 recommendations. For the resulting axial stiffness developing with depth, one
is referred to appendix III.

Figure 6.8: Structural model of pile/jetty supported by Winkler p-y springs + dashpots

The properties of these dashpots will be defined according to the simplified dashpot coefficients that account
for both material damping at the interface and radiation damping in the far field, as obtained from literature
and presented in paragraph 4.6.3.6. The viscous dampers parallel to the static p-y springs can in Seismostruct
only be implemented in the stiffness proportional Rayleigh damping format, resulting essentially frequency de-
pendent damping. This may be assumed reasonable for the radiation damping contribution, but seems unrea-
sonable for the material damping part dominated by hysteretic behaviour. In order to have a best-estimate
damping included in the mathematical dynamic model, it will be fitted at the fundamental frequency of the
structure, that is determined by modal analysis of the system with initial soil stiffness springs included. The re-

63
sulting damping modal damping ratios consequently will be higher for the higher frequency range and lower
for frequencies below the structures fundamental frequency. Structure damping is assumed to have value of
5% critical, which may be assumed conservative for strong level earthquakes. As in Plaxis dynamic analysis a
Hilbert-Hughes-Taylor numerically damped time integration scheme with dissipation parameter α = 0.1 will
be adopted in dynamic Winkler analysis.

As for Plaxis analysis, geometrical nonlinearities will not be included in the Seismostruct Winkler model analy-
sis. As a result second order effects will be neglected. According to Priestley (2007) second order effects signifi-
cance may be assessed by an stability index obtained from drifts, vertical loads and maximum allowable bend-
ing moments:
P∆
θ∆ = (6.3)
Md
Where for steel structures a limit value of 0.05 is proposed. Verification of this criterion for the jetty structure
considered results a maximum allowable drift based on this stability index that exceeds the drift at onset of in-
ground plastic hinging. Hence the prevention of in-ground plastic pile hinging is governing and the P-Δ effects
are not further considered.

The focus in jetty response analysis from the beam on Winkler foundation models will be on:
- Acceleration response in time and frequency domain
- Dynamic peak displacements and (if soil plasticity included) residual displacements
- Pile bending moments and plastic hinging
- Frequency domain representation of soil-structure interaction effects
The obtained response will be compared to the results from simplified dynamic analysis and the results from
coupled 3D Plaxis dynamic soil deposit + jetty structure analysis.

6.6 Coupled transverse dynamic analysis of a slice soil deposit and jetty structure
In this final stage of the study, coupled dynamic analysis of soil deposit overlying engineering bedrock and the
jetty structure is performed.

The bedrock seismic input signals applied to the coupled models are the Duzce and Kocaeli2 recorded accel-
erograms in accordance with paragraph 6.4.1, where bedrock peak accelerations are adopted following from
the probabilistic seismic hazard assessment (see paragraph 5.2):
- ag,peak bedrock = 0.70 g with probability of exceedance 10% in 50 years
- ag,peak bedrock = 1.02 g with probability of exceedance 2% in 50 years

A plane strain 3D finite element model of a slice of the soil deposit including a section of the jetty in transverse
direction is defined. The soil deposit model is defined based on the calibrated model resulting from free field
site response analysis. Nonlinear time history analysis is performed with a damped HHT-Newmark time integra-
tion scheme having dissipation parameter α = 0.1 .

The jetty structure is modelled essentially elastic, where predefined plastic hinges are included at the pile-deck
connection locations according to paragraph 5.5. Possible in ground plastic hinging of the piles is not included
in the models, where this may be justified since from reparability and post-earthquake investigation require-
ments plastic hinging at these locations has to be prevented.

The focus in jetty response analysis from these coupled models will be on:
- Acceleration response in time and frequency domain
- Dynamic peak displacements and (if soil plasticity included) residual displacements
- Pile bending moments and plastic hinging
- Frequency domain representation of soil-structure interaction effects
- Soil failure
A comparison of the obtained response to responses found from simplified dynamic and uncoupled dynamic
design procedures will be made with respect to these response measures.

64
7 Results

7.1 Nonlinear Pushover Analysis

In this paragraph the pushover analysis results of the Winkler and continuum models are presented and dis-
cussed. Initially the load deformation characteristics are determined for a pile loaded at the pile head. Subse-
quently the load deformation characteristics related to the vectorial inertia load pattern based on the systems
eigen frequencies are determined, as is common to form the input for the simplified dynamic analysis proce-
dure.

7.1.1 Pushover Analysis on 2D wall models compared to Winkler pile models

Figure 7.1, Figure 7.2 and Figure 7.3 show the head load-displacement characteristics of the single pile sup-
ported by discrete p-y springs and equivalent walls (EIwall = EIpile) embedded in a soil continuum, for a free head
and a pile head fixed for rotations (as in a jetty structure) respectively.

Figure 7.1: Wall/pile (free head) head displacement ux,head

65
 Figure 7.2: Wall/pile (head fixity) head displacements ux,head

Figure 7.3: Wall/pile (head fixity) head bending moments My,head

The results show the initial stiffness to be in good agreement. Moreover, also the ultimate resistance seems to
correspond relatively well, but the overall stiffness of the pile supported by discrete Winkler p-y springs is
higher. This conclusion may be explained by the wall configuration excluding load spreading/arching effects in
the soil, where for a pile arching effects in the soil decrease effective material stresses and therefore increase
the system stiffness. The results show the response of the Plaxis 2D models to be slightly more flexible than the
Plaxis 3D models. This can be explained by the higher order elements used in Plaxis 2D, resulting is a slightly
more flexible model. Focussing on the composite wall section built from a plate surrounded by a solid, one may
conclude that the behaviour is very close to the behaviour of the equivalent plate. An additional check on the
proper operation of the composite pile was performed by giving the soil an extemely high stiffness, which gives

66
us the pile fixed for translations and rotations at seabed level, which is a case that can easily be checked by the
solution known from elementary mechanics.

In Plaxis 3D, elasto-plasticity for plates is not available, hence an artificial fixation by means of elastic and
elasto-plastic fixed-end anchors was constructed. The proper operation of this solution is proved by the almost
identical load-deformation characteristics and head bending moment development of the Plaxis 2D and Plaxis
3D models as is shown in figures Figure 7.2 and Figure 7.3. Nevertheless for performance based seismic design
purposes the lacking of elasto-plastic plates in Plaxis is a real shortcoming for desiging engineers, and is
strongly recommended for future improvement of the software.

7.1.2 Pushover analysis on 3D single pile models

7.1.2.1 Pushover analysis with derived Hardening soil parameters

For the jetty geometry with pile diameter D=1.4m and a side-side pile spacing of 6m, a 3D soil model width for
a half-pile model of 3m was adopted. Figure 7.4 shows the pushover characteristics (represented by top
displacement and head bending moment development) of the physical 3D pile models in Plaxis 3D, compared
to the discrete Matlock p-y spring models. In contradiction to the wall models discussed previously, in the 3D
pile models, spreading/arching effects in the soil are included resulting a more stiff response compared to the
previously treated wall models.

Figure 7.4: Pile head displacement ux,head for various pile configurations in Plaxis 3D, compared to pile on Winkler nonlin-
ear supports

More stiff behaviour results, and ultimate lateral resistance of the pile in these 3D models is also higher. To
exclude these deviations to be related the finite element mesh configuration, sensitivity of mesh coarseness
and operation of the circular shaped pile are studied by enhancing mesh quality of the circular pile and
calculating the response for an equivalent rectangular pile. The load-displacement results (see Figure 7.4) prove
the to be almost insensitive to model refinements, although mesh quality as indicated by Plaxis was improved
significantly. A similar insensitivity was found for resulting pile head bending moments.

Inclusion of the interface provided some additional flexibility in the Plaxis models, but these effects are
minimized by the program, that determines an optimum between reducing flexibility and preventing ill-
conditioning of het stiffness matrix. However, physical background for the additional flexibility is lacking, since
the interface in the models essentially is a numerical tool. The pile soil interface in Plaxis was modelled as a
rigid interface. Reducing the interface strength parameter Rinter would increase displacements, since more
plasticity at the interface would occur. At the front of the loaded pile this would be nonsense. At the backside

67
of the pile it might be a tool to include gapping, but this turned out to have hardly any effect on global pile
response since very similar pushover response curves resulted. Hence the interface properties are kept default
in order to not converge results of the Plaxis 3D models by just adapting a model parameter without physical
background.

Based on the observations presented above, one may conclude that the stiffer behaviour obtained from Plaxis
3D is a result of the actual behaviour of the Hardening Soil model. Since Matlock p-y expressions are empirically
derived, one would not expect such significant deviations to be found from models in which the soil is
represented by the advanced Plaxis Hardening soil model, although some deviations obviously may be
expected. Hence the Plaxis 3D models need to be validated more precisely in order to be able to judge their
accuracy.

Parameters included in this sensitivity analysis are:

- Plaxis undrained analysis type A (effective stiffness / effective strength parameters) or B
(effective stiffness / undrained shear strength parameters)
- Undrained vs. consolidation analysis for clay
- Over-consolidation of soil layers
- Soil stiffness parameters
Results of variations in these parameters to structural response are discussed next.

7.1.2.2 The type of undrained analysis in Plaxis 3D

Initially, for het clay layers in the model, undrained analysis type A was selected because of generally better
performance of this type of analysis compared to type B undrained analysis. In type A analysis the effective soil
parameters have to be entered, the program subsequently determines the undrained soil behaviour based on
these parameters and the pore pressure generation. Undrained shear strength for these models is a function of
stress state as was already included in the Plaxis Soil Test verifications for the selected parameters.

Since results now seem to deviate significantly from the Winkler p-y curve model resuls, it is investigated
whether these deviations may be introduced by the undrained analysis type in Plaxis 3D. The soil undrained
type of analysis for the clays is set to B, where it should be realized that this removes the stress dependency of
stiffness from the constitutive model and hence actual parameters should be applied. Undrained shear
strength increasing with depth, identical as used for Matlock p-y expressions is included. The load deformation
results of this model results the following conclusions:
- Stiffness is not affected, indicating no significant deviations in plasticity development for the lower
loading range and therefore a proper stress-state dependent inclusion of the undrained shear strength
in the undrained type A soil models.
- Model stability is worse for undrained type B analysis of the clay layers, since no equilibrium can be
found starting already from relatively low load levels. Plaxis 3D indicates that the soil body collapses
which may be doubted, since method A is able to find equilibrium for higher loads.

7.1.2.3 Undrained or consolidation analysis for the clay layers

Initially plastic undrained analysis of clay layers was assumed, because Matlock (1970) p-y expressions were de-
rived for short term loading. Plastic undrained analysis however is equivalent to infinitely short time steps,
which is different from the short time steps in experiments. In order to include effects of this finite time (slight
consolidation possible) in the pushover models, the analysis type is set to consolidation analysis. The perme-
ability coefficients as derived by the soil parameter selection sheet (Appendix II) are used:

Table 7.1: Clay permeability parameters

Soil layer kh [m/s] kv [m/s]
Clay C1 7.5e-10 5e-10
Clay C2 3e-10 2e-10

The intermediate time step for subsequent loading stages is taken 10 sec, which combined with low perme-
ability for clay will hardly result any consolidation, as is the case for the short term behaviour considered.

68
Results of the short term consolidation pushover analysis are presented in Figure 7.5, where they are plotted
together with the load-displacement characteristics of the model with undrained constitutive clay material be-
haviour and the discrete p-y soil model.

Figure 7.5: Pile head displacement ux,head for different Plaxis undrained analysis types

Figure 7.5 results two conclusions:

- Accepting limited consolidation for the clay material does not affect the global response characteris-
tics of the pile-soil system
- The consolidation analysis model in Plaxis 3D is more stable for tracing the load-displacement path be-
yond onset of plastic hinging at the pile head, were a sudden change in soil state will result. A closed
explanation for this remarkable observation has not been found, but it appears to be related to local
instability resulting an unacceptable error and therefore no convergence of the finite element solu-
tion. The ability of tracing the load displacement path in the very nonlinear range is interesting from
the displacement based design perspective, where ductility of the system has to be proved. It is there-
fore recommended to improve the stability of plastic analysis in Plaxis.

7.1.2.4 Overconsolidation of the soft soil layers

Based on correlations with unit weight, void ratio and undrained shear strength, the clay layers present appear
to be slightly over-consolidated. Consequently an OCR of 1.5 was adopted in the models. However, for clay it is
known that OCR values based on these correlations may actually not be observed if aging effects can also affect
soil properties like water content and plasticity index. For the slightly over-consolidated parameters, the consti-
tutive behaviour of the soil according to the Hardening Soil model, is dominated by the more stiff elastic
unload/reload behaviour. The p-y expressions according to Matlock were derived for normally consolidated or
slightly over-consolidated soft clays. Although an OCR equal to 1.5 is relatively low, its effects on global struc-
tural response are considered in this study. Material parameters for both clay layers were set to normally con-
solidated and effect is presented in Figure 7.6.

69
 Figure 7.6: Sensitivity of pushover curve to clay over-consolidation

The figure shows the influence of the over-consolidation on global pushover response characteristics. The fol-
lowing conclusion can be drawn:
- The effects of over-consolidation are clear, but can be considered relatively limited. Stiffness decrease
is in the order of a 10% for the normally consolidated compared to the over-consolidated (OCR = 1.5)
clay material state.

This can be explained by Figure 7.7, that shows the very local concentration of stresses, that exceed the pre-
consolidation pressure already for relatively low lateral loads at the pile head. Beyond the pre-consolidation
pressure the soil material stiffness decreases, resulting an increasing contribution of local normally consoli-
dated soil state to the global lateral pile response.

Figure 7.7: Overconsolidation ratio development for increasing lateral pile loads (0, 100, 200 kN/pile)

70
It is noted that the change from over-consolidated soil state to normally consolidated soil state reduces the ef-
fective undrained shear strength, which is a function of local stress state, as can be observed from tri-axial
tests. In this sense it is inconsistent to compare the Plaxis 3D outcome of models with normally consolidated
clay to the p-y models that are based on cu values for the slightly over-consolidated clays.

7.1.2.5 Soil stiffness parameters

Soil stiffness parameters initially included in the models are best estimates based on available soil data and cor-
relations from literature. Nevertheless, stiffness parameters for clay are in general very difficult to classify
when limited soil survey is available. Even stiffness parameters resulting from tests always should be treated
carefully, as the handling of the soil may significantly disturb local stress state and especially soil stiffness pa-
rameters. As a result, it is common in geotechnical engineering to multiply/divide soil stiffness parameters by a
factor 2 in order to establish sensitivity of the analysis to variations in these stiffness parameters.

In the present study, the soil stiffness parameters are divided by a factor 2, in order to investigate whether this
may reduce global stiffness towards the lower stiffness as is observed for pile models with soil represented by
the Matlock p-y expressions. The results are shown in Figure 7.8

Figure 7.8: Sensitivity of pushover curve to clay stiffness variations

Based on Figure 7.8 the following may be concluded:

- The reduction of initial stiffness for the global pile-soil system is order 10%, which is much lower than
the soil stiffness reduction of 50%. Hence, the soil response seems to be strongly affected by local
plasticity in the soil adjacent to the pile and stiffness dependent deformations of a larger soil body
seems to have a minor effect. The reduction percentage is comparable to the initial stiffness reduction
resulting if the clay layers are given a normally consolidated state.
- Even when the best-estimate stiffness parameters are reduced by a factor 2, the global response is
more stiff than the response obtained from the pile-Winkler p-y soil models.

7.1.2.6 P-y expressions for the discrete Winkler foundation after Jeanjean. (2009)
Based on the previous considerations, a more stiff response is obtained for the Plaxis 3D pile soil models com-
pared to the models with discrete p-y soft clay soil after Matlock, irrespective of any reasonable variations in
soil parameters in the Plaxis 3D models. Consequently it is doubted whether Matlock’s p-y expressions are rep-
resentative for soft clays.

Although Matlock’s expressions are most commonly applied in design, several studies (Martin & Randolph,
2006; Jeanjean, 2009) have shown that they may generally be too conservative. Alternative p-y expressions
with test verification have been proposed, of which the data sets from Jeanjean (2009) fit best to the case con-

71
sidered in this study, in which soil response is dominated by the very soft to soft clay deposits. Jeanjean (2009)
proposes p-y expressions that fit in the equation of (O'Neill, Reese, & Cox, 1990):

 G  y 0.5 
p = pu tanh  0    (7.1)
 100cu  D  
 
Where:
pu = N p cu
 −ξ z 
 
N p = 12 − 4e D 

0.25 + 0.05λ ( for λ < 6)

ξ=
0.55( for λ ≥ 6)
cu , seabed
λ=
cu ,incr D
Where pu is a function of undrained shear strength, pile diameter and depth.
Jeanjean showed in his study that the Matlock p-y expressions are significantly underestimating both modulus
and ultimate resistance of soft clay soil. The tests performed by Jeanjean to verified the proposed p-y expres-
sions were performed on a 1:48 scale, which might be a source for deviations when using these expressions for
big diameter piles.

As for the Matlock (1970) p-y expressions, the actual p-y curves according to Jeanjean as a function of depth
and estimated soil parameters are determined in Excel (Appendix III) Again for practical convenience bilinear
fittings were sought, which subsequently were adopted in the Seismostruct pushover analysis.

The global response results are presented in Figure 7.9.

Figure 7.9: Performance of conventional Matlock (1970) p-y curves and recently developed Jeanjean (2009) p-y curves for
soft clay compared to 3D Plaxis modelling

As shown in the figure, the Plaxis 3D model and Seismostruct model with Jeanjean p-y springs result global
lateral pile responses that are very similar. In order to further investigate the local soil actions on the pile for
both models, the development of bending moment along the pile are plotted in Figure 7.10

72
 Figure 7.10: Local bending moment along pile for discrete nonlinear Winkler pile model and Plaxis 3D pile model

The conclusions that can be drawn from Figure 7.9 and Figure 7.10 are:
- The global response resulting from the model with p-y soil after Jeanjean (2009) fits well to the global
pushover response obtained from the Plaxis 3D model with the soil parameters most likely based on
the parameter sheet.
- The local soil actions on the pile resulting from the Plaxis 3D analysis fit to the local soil actions of the
Jeanjean p-y springs, providing confidence in accuracy of both analysis. Only slight deviations in soil
response are found for the deeper clay layers, that in the Plaxis 3D model behave slightly more stiff,
resulting in a stronger decrease om the pile bending moment with depth in these soil layers.
- The gradient of the bending moment line near pile tip level is relatively low, especially for the lower
loading. Consequenly shear force in the pile is low at this level, indicating the limited influence of the
presence of the stiff sand layer on lateral pile response. This was also to be expected based on the
0.25
medium-long characterisation of the pile (λ = (ksecant,soil/EIpile) ) according to the beam on winkler
foundation concept, which implies that the tip boundary condition is of limited importance for pile
head response and the lateral pile behaviour to be governed by the soil response in the top soil layers.
- Although most common and adopted in API and ISO design guidelines, Matlock p-y expressions (as
also reported in literature) appear to be too conservative.
- The p-y expressoins after Jeanjean give approximately similar response as the Plaxis 3D model with the
advance hardening soil model. This supports the Jeanjean expressions to give reasonable results, even
when extrapolated to diameters much bigger than the diameters used in the tests performed by
Jeanjean that are at the basis of the proposed p-y expressions.
- Based on Jeanjean (2009) and the present study, it seems reasonable to use the Jeanjean p-y
expressions for future design applications. However, despite the good fit for the current project,
verification by a Plaxis 3D analysis is to be recommended because of the well known limitations of the
Winkler model concept. In this study p-y expressions according to Jeanjean will be adopted. Dynamic
stiffness multipliers according to El Naggar and Bentley (2000) will not be applied in dynamic analysis,
since they were derived for the much more flexible Matlock static p-y springs and consequently may
lead to an overestimation of dynamic stiffness.
- Despite the locally nonlinear soil response as present in the analysis, global soil response is almost
linear. Consequently also structure global response is approximately linear up to the onset of plastic
hinging at the pile head. Then in the second branch up to failure, again almost linear behaviour is
obtained. Consequently, an equivalent linear system may be determined for calculating response of
pile-soil systems for low level earthquakes, however for higher intensity earthquakes structural

73
 nonlinearities and soil failure are expected to have a pronounced effect on the response and should
be included in the analysis.
- The bilinear elastic - perfect plastic approximation of global system response as proposed in the Fajfar
(1999) simplified dynamic analysis method does not account for post-yield hardening in the system,
which based on the pushover results seems to give a conservative approach for the pile-soil system
that shows significant increase of resistance (±40%) in the “post-yield” branch.

7.1.3 Linearization of pile-soil interaction

The approximate bilinear load deformation relationship as is obtained from Figure 7.9 for both the discrete
Winkler p-y model and the finite element continuum model imply that the system behaves almost linear for
lateral loads that are below the load level corresponding to onset of plastic hinging in the pile. Despite the very
local nonlinear characteristics of the soil structure interaction, it must be concluded that averaging over the
pile length results an approximately linear system. This important conclusion can be used for calculating pile-
responses for low level earthquakes by means of linear dynamic methods which is very beneficial from compu-
tational point of view. From this conclusion the question arises which lateral pile-soil spring stiffness to apply in
linear models. Lateral pile deformations (except from the very top part of the pile) remain relatively low up to
onset of plastic hinging. Hence it seems reasonable to adapt the upper bound initial bilinear spring stiffness
(which was used as an approximation of actual p-y curves) for linear low load level calculations. The result is
shown in Figure 7.11.

Figure 7.11: Effect of linearized soil springs on lateral pile response

A slightly stiffer response is obtained from the linear springs, but it may be concluded that a good approxima-
tion of global lateral pile response can be found by the linear approximation up to the load level corresponding
to the onset of pile plastic hinging. However for high level earthquakes, according to performance based design
principles, deformation capacity of the structure should be used to dissipate energy. As a consequence for high
level earthquake design a nonlinear system should be considered.

7.1.4 Equivalent Plaxis 2D model to represent lateral single pile response

In order to study the arching/spreading effects in Plaxis, Plaxis 3D analyses with models of different width have
been performed, implying a pile having a particular side-side spacing from adjacent piles.
The results are shown in Figure 7.12

74
 Figure 7.12: Lateral pile response sensitivity to pile side-side c.t.c. spacing

Figure 7.12 shows that based on the Plaxis 3D analyses, no efficiency reduction for side-side spaced piles has to
be expected in the soft soil conditions for a pile spacing exceeding 2.5D, and hardly any reduction is observed
already beyond a spacing of 2D.

According to paragraph 4.6.3.9 and the Plaxis 3D and 2D results regarding the influence of side-side pile spacing
one may conclude the following:
- Performing 2D analysis on a pile (wall) model having the actual pile stiffness and a soil slice width of
1m, underestimates the stiffness and capacity of the soil when the pile spacing exceeds 1D.
- As a consequence one needs to adapt pile or soil properties in order to have a proper relative stiffness
distribution in the model.
- Adapting the soil parameters is not very straight-forward, since both stiffness and strength properties
of the soil in the hardening soil model will affect global strength and stiffness properties of the system
in a truly non-trivial way.
- Consequently the best options available seems to reduce the pile stiffness properties and the lateral
load (pushover forces as mass times acceleration), and then soil properties can remain unchanged. It is
noted that this approach may be suitable for pushover analysis and dynamic analysis of head loaded
piles (or inertial pile loading). It however will not be convenient for dynamic analysis of soil deposit +
structure, where lateral pile load cannot be reduced since it is formed by kinematic pile loading of soil
deposit responses for an applied bedrock signal.

The following figure shows the load deformation characteristic of a single pile, loaded by a lateral concentrated
load at the head. From the figure one may conclude that reasonably well estimates of the actual (3D) pile be-
haviour are obtained from the equivalent 2D analysis when both pile capacity and load are reduced by the side-
side pile spacing. Furthermore, one may conclude that reduction of the pile properties and load by a factor ex-
ceeding 2.5D is not to be advised (even when the pile spacing exceeds this spacing distance), since this would
overestimate the relative soil stiffness and strength capacity in the models.

75
 Figure 7.13: Equivalent mobilized soil width dependence on side-side pile spacing

Figure 7.14 shows the head bending moment of a single pile as a function of lateral loading for the various
equivalent models related to varying pile spacing. From the figure one may conclude that the head bending
moment is relatively insensitive to the pile spacing and head bending moments are reasonably well repre-
sented by the equivalent 2D models.

Figure 7.14: Pile head bending moment My,head sensitivity to pile spacing

From the previous it may be concluded that jetty lateral pile load-deformation characteristics may reasonably
well be approximated by adopting a 2D equivalent model. For soft soil conditions, as present in the considered
case project pile spacing has hardly any effect on pile efficiency if the side-side spacing exceeds 2.5D, and con-
sequently the reduction factor for pile stiffness, pile top fixation and lateral loading should not be chosen
higher than 2.5D.

7.1.5 Pushover analysis of jetty cross-section (3 rows of piles in loading direction)

This paragraph presents the results from pushover analysis on the jetty cross-section as was presented in para-
graph 5.1. The outcomes of pushover analysis from 3D Plaxis is assumed to represent reality reasonably well, as
was proved in section 7.1.2 where the modelling of a single pile response in Plaxis 3D has been proved to fit

76
reasonably well to test results from Jeanjean (2009). In this section the effects of grouping in a jetty transverse
configuration are studied, and appropriateness of defining an equivalent 2D model is investigated.

Jetty global response in this section is again studied by comparing the lateral deck displacement and the fixa-
tion bending moment at the connection of the pile to the deck, which are typical representative variables for
global response in performance based design. The results for these variables with increasing pushover loading
are presented in figures Figure 7.15 and Figure 7.16. As shown in Figure 7.15 a reasonable equivalent 2D model
can be found. As for the single piles a considerable reduction of structure stiffness and pushover loads however
is required, even for pile groups with piles spaced in the loading direction. These required reductions make the
approach is not suitable for performing seismic analysis of piles embedded in a soil deposit, since it would not
be convenient to reduce bedrock loads. On the contrary when keeping the actual seismic loads, a lower pile
stiffness would result unrealistic displacements and including the actual pile stiffness would overestimate pile
stiffness compared to soil stiffness.

Figure 7.15: Pushover characteristics for jetty cross-section with piles spaced in loading direction, sensitivity to increased
relative soil stiffness

Figure 7.16: Pile head bending moment My,head development for

77
Based on the analyses results, the following conclusion can be drawn:
- A significant efficiency reduction is obtained due to the pile interaction in the loading direction. Ac-
cording to Reese and van Ympe (2001) the efficiency factor is 1.00 for the leading piles and 0.79 for
the trailing piles (see appendix III), which would result in an average efficiency of 0.86 for a jetty trans-
verse pile row of 3 piles. The system lateral efficiency obtained from the Plaxis 3D analysis for the 3 in
row jetty piles is ranging from 0.87 to 0.65 with increasing load level compared to single pile lateral
stiffness. The lateral system response however is affected by both structure stiffness and soil stiffness.
Consequently soil subgrade reaction efficiency for the grouped piles will be even lower than these ef-
ficiency factors. Actual soil efficiency was found by reducing the p-y expressions initial stiffness and ul-
timate capacity up to a level for which again a good fit is found with Plaxis 3D as there was for single
pile response. The average efficiency was found to be approximately 0.55 for the leading and two trail-
ing piles. Apparently the efficiency reduction is higher than provided by Reese and van Ympe and is
more in accordance with the higher efficiency reductions as proposed by NCHRP report 461
(Transportation Research Board, 2001). The significant reduction of pile efficiency for the trailing piles
is explained by the inability of the soil to mobilize lateral resistance as is indicated by Figure 7.17.
(horizontal section at z = -23.5m+MSL) that shows relatively low soil strains in the zone between the
pile implying a kind of rigid body movement of the local soil.

Figure 7.17: Relatively low Cartesian strains εxx in the soil between the piles for Plaxis 3D jetty modelling in soft soil

Zooming in at the local stress development in the soil in between the pile however reveals the load spreading
effects in the soil in Plaxis 3D also for the trailing piles, as becomes clear from comparison of Figure 7.18 and
Figure 7.19. For the 2D equivalent model soil stresses mobilized by the displacement of the trailing piles (walls)
are perfectly passed through to the leading pile (wall), and real passive resistance is mobilized in front of the
leading pile. For the 3D model however mobilized resistance is to some extend spread over a soil width exceed-
ing the pile diameter, which results a decreasing stress with increasing distance from the pile towards it leading
pile at the piles centre plane y = 0 m.

78
 Figure 7.18: Cartesian total stress σxx for Plaxis 3D jetty modelling

Figure 7.19: Cartesian total stress σxx for equivalent Plaxis 2D jetty modelling

It is concluded that an equivalent Plaxis 2D model may be defined for pushover analysis of a jetty structure.
Since computationally effort of pushover analysis is limited this however would not be very useful. , where this
equivalent 2D approach might be attractive to perform efficient dynamic analysis of piles + soil subjected to
dynamic pile head loading, as for example dynamic analysis of piles under offshore wind turbines or for piles as

79
machine foundations. Hereby it is noted that one should make sure that the dynamic pile characteristics are in-
cluded correct and an appropriate Winkler or Plaxis 3D continuum model conceptually would be preferable.

Aiming at the analysis of pile-soil interaction for jetties under seismic loading, the equivalent Plaxis 2D model is
not suitable, since load is formed by a bedrock excitation which cannot be reduced inherently with pile stiffness
since this would underestimate soil deposit response levels. Hence the dynamic jetty slice transverse response
for earthquake signals preferably is to be analyzed with a dynamic structure on Winkler foundation model or by
a coupled analysis in Plaxis 3D dynamics module, that show very similar pushover load-deformation character-
istics as presented in Figure 7.20. and Figure 7.21.

Figure 7.20: Pushover curve jetty loaded by force vector proportional to the fundamental mode vector

Figure 7.21: Pile head bending moment development for jetty pushover loaded by force vector proportional to the fun-
damental mode vector

For the case jetty configuration an average efficiency of the Winkler foundation was found to be 0.55ver for the
row of three piles. For the Plaxis 3D model a width of 2.5*D was calibrated to be sufficient for the piles in soft-
soil conditions.

80
Based the results obtained, one may conclude that by applying the simple Winkler foundation concept with p-y
curves as prescribed in the preceding, the monotonic load-deformation characteristics found from advanced fi-
nite element analysis can be approximated quite well. For cyclic behaviour however one may expect the re-
sponse from Winkler analysis to deviate somewhat from Plaxis finite element analysis, since soil deformations
and soil state will not be completely restored in Plaxis when monotonically loaded up to a level beyond the on-
set of plasticity and subsequently unloading and reloading in opposite direction. These deviations were investi-
gated and the result is shown in Figure 7.22.

Figure 7.22: Cyclic pushover characteristics from Winkler p-y and Plaxis 3D HS analysis

Clearly a preference direction of the structure deformation response is developed in the Plaxis 3D HS analysis,
where this effect is not present in Winkler p-y analysis but will be present in reality for loading considerably
into the post-elastic range. Plaxis pile-soil interfaces however do not properly include gapping closure and
hence the preference direction found from these analysis may be somewhat overestimate reality. It is noted
that the development of a preference direction of displacements will be even more pronounced for pile-deck
structures founded in slopes, for which the simplified Winkler p-y analysis hence will introduce significant er-
rors when not accounted for.

81
7.2 Free field site response analysis

7.2.1 Calibration of the finite element model

7.2.1.1 Plaxis finite element mesh size

Initially the element size applied in the numerical analyses was based on references as noted in paragraph
6.4.4.1. For this mesh configuration analysis are performed for harmonic and random bedrock signals. The re-
sults show that a very good fit of the numerical finite element and frequency domain analysis responses is ob-
tained. The very fine meshes however are not convenient with respect to computational requirements (espe-
cially towards 3D FE models) and therefore it was investigated how responses are affected by applying some-
what bigger elements. It is expected that resulting errors are bigger for high frequency components of the sig-
nal, since related wave lengths for these frequencies are smaller. However, since the high frequencies are ex-
pected to be of minor importance for resulting forces and deformations, some error of resulting response for
this frequency range may be accepted.

The effects of increasing element sizes by approximately a factor 2 are presented below for both the 3 and 10
Hz harmonic signals and linear elastic material with 5% critical added Rayleigh viscous damping in Figure 7.23
and Figure 7.24.

Figure 7.23: Sensitivity of acceleration response to element size for 3 Hz harmonic bedrock excitation

Figure 7.24: Sensitivity of acceleration response to element size for 10 Hz harmonic bedrock excitation

82
From the figures it can be concluded that the bigger mesh element does not affect the 3 Hz harmonic input sig-
nal response and for the 10 Hz harmonic input signal a reduction of peak accelerations of approximately 10% is
obtained. Consequently the error resulting for natural seismic signals will be depending on the frequency con-
tent of the input signal. Actual signals have the highest energy content generally in the frequency range 1-10
Hz, which would imply that errors in accelerations due to a factor two increased element sizes can be esti-
mated to be <5% non-conservative.

The figures above show the clear effects mesh coarsening can have. However, with respect to computational
requirement lowering the number of elements has appeared to be critical. Hence, in this study it was decided
to accept the slight deviations for the higher frequencies, by adopting the somewhat coarser mesh compared
to the recommended maximum element sizes (Lysmer & Kuhlmeyer R.L., 1969). Sensitivity of results to further
coarsening of the mesh is not included in this study but probably not to be recommended with respect to accu-
racy of the results.

7.2.1.2 Finite element model width

Based on bedrock excitations by both the random Duzce signal and the simple harmonic signals the model
width was calibrated, focussing on the two criteria as noted in paragraph 6.4.4.2. The sufficient model width
was found to be 300 m, with standard Plaxis viscous boundaries applied at the xmin and xmax boundaries. Results
supporting this conclusion are presented below.

With respect to horizontal seabed accelerations (first criterion paragraph 6.4.4.2) it was found that the bound-
ary effect reach approximately 50 to 100 m away from the boundary. At 100 m from the boundary sufficient
convergence of horizontal accelerations is obtained. Considering the time accelerations history obtained one
may conclude that model width variations seem to most affect the obtained horizontal accelerations in the lat-
est stadium of the earthquake, but for the 200 m model also clear deviations of peak accelerations (up to 25%
conservative and non-conservative) in the strong motion stage are obtained. Apparently wave reflections are
becoming significant for the centre model response and 200 m model width is insufficient.

With respect to vertical accelerations at seabed level (second criterion paragraph 6.4.4.2), no real convergence
2
is obtained towards the widest models, but clearly very low acceleration levels (up to approximately 0.1 m/s )
are obtained at the model centre seabed for all three model widths. Away from the model centre, vertical ac-
celerations grow rapidly, especially for the smaller models as is shown in Figure 7.25 for location (sea-
bed,x=50m)

Figure 7.25: Vertical seabed accelerations at x=50m for model width 400, 300 and 200 m

When subjected to the 3 Hz harmonic input signal, peak vertical accelerations in the centre region are about
2
0.04 m/s , which is less than 1% of the horizontal peak accelerations. Towards the boundaries the peak vertical
2
accelerations grow up to a maximum absolute value of about 0.4 m/s . For the 10 Hz harmonic bedrock signal

83
 2
peak vertical accelerations are below 0.02 m/s which is smaller than 2% of the obtained peak horizontal accel-
2
erations. For this signal the peak vertical acceleration found towards the lateral boundaries is about 0.3 m/s .

The vertical oscillations in the 2D Plaxis models are dominated by wave reflections at the boundaries generat-
ing surface waves. Analyses have shown that the effects are significant in the area near the lateral boundaries
for both the random and harmonic bedrock motions. However, vertical accelerations are very limited in the
centre of the model already for a critical damping percentage of only 5%, which is a lower bound approxima-
tion for actual soil damping levels. It may therefore be concluded that boundary effects are sufficiently damped
out towards the model centre for the 300 m model and consequently this model width will be used in the next
steps of this study. In the 1D frequency domain analysis vertical motions are not considered and the response
for an infinitely extending soil column subjected to uniform bedrock excitation is determined, hence it cannot
be used as a comparison in this sense.

7.2.2 Comparison of responses from the linear elastic finite element model and equiva-
lent linear frequency domain analysis for layered soil profile
The finite element acceleration responses obtained show a proper fit with the linear frequency domain solu-
tion for both harmonic bedrock excitation and random seismic bedrock signals. The applied harmonic signal
was smoothened to avoid high frequency noise resulting from numerical problems in Plaxis with respect to ini-
tial conditions. Plaxis has shown to be very sensitive for both initial conditions and phase-to-phase transitions
in calculations, which is a problem recommended for future Plaxis improvement and in this study was solved by
inclusion of a numerically damped time integration scheme. Figure 7.26 shows the almost perfect fit for the 3
Hz harmonic input signal in the time domain. A conservative soil material damping ratio of 0.05 was adopted in
this analysis.

Figure 7.26: Excitation and response acceleration time histories for a 3 Hz harmonic signal

For the high level earthquakes commonly much higher damping levels are realistic, especially for soft soil con-
ditions. This much higher material damping results from material internal hysteretic behaviour, even for load
levels below failure. Linear elastic and Mohr-Coulomb soil models do not include this hysteretic damping in the
elastic range, that on the contrary is captured in the HSsmall soil model. Consequently for LE and MC soils an
equivalent damping has to be determine preliminary, for which in this study the equivalent linear frequency
domain analysis was used for this because of computational convenience. The resulting effective soil parame-
ters for a peak bedrock acceleration of 0.5g are as presented in Table 7.2.

84
Table 7.2: Effective soil dynamic parameters based on equivalent linear frequency domain analysis
Soil layer Clay C1 Clay C2 Sand S3
ϒeff [-] 0.0026 0.0011 0.0006
Geff [kPa] 4790 23790 126250
2
Vs,eff [m/s ] 53 116 248
ξeff [-] 0.179 0.093 0.063

Applying these parameters in linear elastic finite element analysis should essentially give results close to the re-
sults from frequency domain analysis. However, when adding Rayleigh damping to linear elastic material the
target damping ratio cannot exactly be met over the entire frequency range. A best-fit Rayleigh damping was
applied for the frequency range of interest, which for the Duzce signal combined with the soil profile is ap-
proximately 0.1 to 10 Hz.

Table 7.3: Rayleigh critical damping ratios estimated based on equivalent linear frequency domain analysis
Damping Best estimate Rayleigh damping input frequencies
Soil layer f1 [Hz] ξ1 [-] f2 [Hz] ξ2 [-]
C1 clay 2 0.20 7 0.20
C2 clay 2 0.10 7 0.10
S3 sand 2 0.07 7 0.07

The seabed horizontal acceleration response spectra of both equivalent linear frequency domain analysis and
linear elastic and Mohr-Coulomb finite element analysis are presented in Figure 7.27.

Figure 7.27: FFT amplitudes of seabed horizontal acceleration response due to Duzce bedrock signal

The figure shows the good fit of the Plaxis finite element model seabed response compared to the equivalent
response by the equivalent linear frequency domain analysis. A somewhat higher spectral response of the finite
element analysis is obtained in the frequency range around 3 Hz, which is related to the slightly lower Rayleigh
damping level for these frequencies. Similar reasoning applies for the slightly lower frequency content of the
finite element response around 1 Hz.

85
7.2.3 Site response for linear elastic, Mohr-Coulomb, HS and HSsmall soil constitutive
models

7.2.3.1 Comparison of responses for linear elastic, Mohr-Coulomb and HSsmall soil consti-
tutive models
In this paragraph the sensitivity of the site response to variation of the soil constitutive model is discussed. Four
constitutive models are considered, being the linear elastic, Mohr-Coulomb, Hardening Soil and Hardening Soil
with Small strain stiffness models (Plaxis Material Models Manual, 2011). Bedrock peak acceleration in this
study was taken to be 0.5g, which is considered a common level representative for strong earthquakes. The
linear elastic finite element analysis have shown to match the frequency domain solution very well. However,
nonlinear physical phenomena like hardening/softening, plasticity, stress dependency of stiffness and stiffness
dependency on loading history are not included in these analysis, where they are obtained for real soils. Con-
sequently one may expect the Plaxis advanced soil models to result more accurate site responses, provided a
good numerical performance.

The seabed horizontal acceleration Fourier response spectra for the LE, MC and HSsmall soil models are pre-
sented in Figure 7.28.

Figure 7.28: FFT amplitudes of seabed horizontal accelerations for the LE, MC and HSsmall soil models, subjected to
Duzce bedrock signal

The figure shows the almost similar spectral response for the LE and MC soil models. The minor differences ob-
tained are the somewhat lower spectral peaks for the MC model and the somewhat higher frequency content
at the higher frequencies for the MC soil. The somewhat lower spectral peaks are explained by Figure 7.29 that
shows the clear effects plasticity have on development of peak accelerations.

86
Figure 7.29: Mohr-Coulomb plasticity affecting free field accelerations, response subinterval 5-6 s for the Duzce signal

For the HSsmall however a significantly different response spectrum is obtained (Figure 7.28). Spectral peaks
around the high frequency range of the input signal (~3 Hz) are much lower for the HSsmall model, where it has
a much higher frequency content for the lower and higher frequency range. According to Brinkgreve, Bonnier
and Kappert (2007) and oral communication with R.B.J. Brinkgreve, Nov 2011, 1-2% Rayleigh damping was
added to the HSsmall material, in order to account for energy dissipation at very low strain levels as is observed
in reality and not covered by the hysteretic damping included in the HSsmall model. This low Rayleigh damping
percentage however is much lower than the Rayleigh damping added to the LE and MC soil models. Conse-
quently much higher total damping is included in the LE and MC models for the lower and higher frequency
range, which partly explains the higher frequency content of the HSsmall response for these frequency ranges.
However, the total spectral response obtained for the HSsmall model is relatively far from the LE and MC re-
sponse, for which reason it was decided to assess the HSsmall model response in more detail as presented in
the next paragraph.

More insight regarding the response for different soil constitutive models is provided by analyzing the accelera-
tion responses in the time domain. The acceleration time history for seabed level due to the Duzce random
bedrock signal for time interval 0-10 sec is presented in Figure 7.30

Figure 7.30: Horizontal acceleration response time histories (subinterval 0<t<10 s) at seabed level due to bedrock excita-
tion for Duzce bedrock acceleration signal

87
For very low strain levels the HSsmall model underestimates soil damping levels (Brinkgreve, Bonnier and Kap-
pert (2007), which may contribute to the higher response for the initial phase of the earthquake. The 1-2%
Rayleigh damping for the frequency range of interest (0.5 – 8 Hz) was included in two different ways, mass-
and-stiffness proportional and stiffness proportional only. The first option results very high effective damping
levels for the very low frequency range, which for the Duzce signal and present soil turned out to be not very
significant. However an appropriate definition of the additional Rayleigh damping should be made case de-
pendent based on input signal frequency content and system fundamental frequency level.
The material viscous damping percentage in equivalent linear frequency domain analysis and linear elastic and
Mohr-Coulomb finite element analysis is taken constant over the total time history. The damping level is de-
termined as a function of the extreme shear stress level, which generally will overestimate realistic damping for
low level response phases and underestimate damping for high level response phases. These effects become
clear from Figure 7.30, showing the acceleration amplitudes developing over the time interval. A more realistic
approximation of soil responses by linear elastic analysis may be obtained when sub-phases with varying damp-
ing levels are defined, which however doesn’t fit in the Fourier frequency domain analysis and cannot be ap-
plied in dynamic Plaxis analysis method, since for each sub-phase start strong deviations of response will result.
It may be clear from these conclusions that properly accounting for soil damping by adding of Rayleigh damp-
ing is difficult and so the HSsmall model feature of including hysteretic damping is very attractive from this per-
spective.

7.2.3.2 Performance of the HSsmall soil constitutive model in dynamics

For strong motion phases of the time signal, the HSsmall material site response shows a phase delay compared
to the linear elastic and Mohr-Coulomb materials. This out of phase response is not obtained in the initial low
intensity interval and also is restored when motion levels decrease. Since this effect is not present for Mohr-
Coulomb materials it cannot be related to plasticity and it needs to be a result of the stress dependent stiffness
development in the HSsmall model. Furthermore sudden acceleration jumps in response of HSsmall modelled
clay layers is obtained, which seems not very realistic when the soil responses at bedrock, layer interfaces and
seabed levels are presented for a shorter time interval as presented in Figure 7.31.

Figure 7.31: Horizontal soil accelerations at various depths

These sudden accelerations jumps are not obtained for the HS , Mohr-Coulomb and linear elastic constitutive
soil models. The most sudden changes in accelerations are obtained at seabed level, but also the deeper C1-C2
layer interface shows a very unrealistic acceleration response. However, Figure 7.31 a more smooth and realis-
tic response for the sand layer underneath. Hence it may be concluded that the HSsmall soil constitutive model
is not resulting unrealistic responses in general when applied in dynamics. It seems that just a peculiar configu-
ration of model layering, soil parameters and the HSsmall model operation in dynamics results the sudden
jumps in acceleration level, which is considered in more detail.

The sudden changes in the soil accelerograms were found to be a consequence of the initial very high small
strain stiffness when the small strain overlay model resets stiffness. Reference in this explanation can be made

88
to the free vibration equation of motion in dynamics (equation ), where for simplicity damping related terms
are neglected:
[ M ] u&& + [ K ] u = 0 (7.2)
Reset of the small strain stiffness gives an abrupt change in stiffness matrix [ K ] , since [ M ] is constant and
u has to be continuous, accelerations will abruptly develop too, in order to obey the equation of motion.
The HSsmall model returns to the small strain stiffness when according to Benz (2006) the scalar valued shear
strain:
H ∆e
γ hist = 3 (7.3)
∆e
is reset to zero by the deviatoric strain history tensor H that multiplies the deviatoric stain increment ∆e in
the numerator, which occurs when a deviatoric principal strain rate reversal is detected. Apparently the sudden
changes in stiffness (high ratio G0 / Gur then for soft soils) strongly affects the accelerograms, resulting the
sudden jumps in accelerations. The direct link between principal strain rate reversal and sudden change in ac-
celeration level is demonstrated by Figure 7.32 and Figure 7.33 that show horizontal accelerations and principal
strains for the locations (0,-23) and (0,-33) in the C1 and C2 clay layers for the Duzce seismic signal (5-6 sec).

Figure 7.32: Horizontal accelerations for the C1 and C2 clay mid-layer levels

Figure 7.33: Principal strains for the C1 and C2 clay mid-layer levels

89
Sensitivity of the acceleration for the following analysis input parameters was considered:
- Number of dynamic sub steps
- Undrained/drained analysis of the clays
- Small strain overlay model parameters G0 and ϒ0.7
- Additional Rayleigh damping included in the HSsmall model
- Stress dependency included in HSsmall soil model

The number of dynamic sub steps as determined in 6.4.4.1 is significantly lower than the default number of dy-
namic sub steps provided by Plaxis. This is caused because Plaxis calculated the number based on the compres-
sion wave velocity that tends to infinity when undrained conditions are applied to soils. Since the problem
studied is expected to be dominated by shear loading a lower number of dynamic sub steps was chosen for
computational convenience. In order to exclude a relation of the unrealistic sudden jumps in acceleration ob-
tained now with the lower number of sub steps an analysis was performed with a higher number of dynamic
sub steps. Results of this analysis still show the sudden jumps in acceleration.

Initially undrained analysis type A was applied for the clay layers, as has proven to give good results and stable
calculations in static analysis. However, the unrealistic acceleration response in dynamic analysis was obtained
for the undrained clays and not for the drained sand. Hence analyses have been performed in which drained
analysis was applied to the clays as well. It was found that the only a slight smoothening of the blocked shape
accelerograms in the clay is obtained and no significantly different response is obtained as is reasonable for this
problem dominated by shear loading. Besides, as was noted before, the MC acceleration responses show
lower peaks than the LE, which was explained by the effects of plasticity. It is noted that in Undrained analysis
type A higher deviatoric stress levels can be accepted for MC soils compared to HS soils, since pore pressures in
the HS and HSsmall models result a decrease of effective isotropic stress and consequently a lower Mohr-
Coulomb failure limit. If this feature explains the much lower acceleration response peaks for the HSsmall soil
was analyzed by comparing the results of drained and undrained type A analysis for the clays. It was concluded
that an almost similar response was obtained, so pore pressure generation cannot be considered governing in
this sense.

According to various empirical relations (e.g. (Alpan, 1970) and others) the ratio G0 / Gur varies with soil type.
In general a higher ratio is obtained for clay than for sand. In this study the selected ratios are 6.9, 6.9 and 1.8
for the C1 clay, C2 clay and S3 sand layer respectively. Values applied for ϒ0.7 are based on empirical relation-
ships (Benz, 2006; Stokoe et al, 2004) were it is considered a function of plasticity index and overconsolidation
ratio for clays and a function of estimated relative density for sands. Since the clays typically have a
higher G0 / Gur ratio and threshold shear strain ϒ0.7 compared to sand, it was doubted whether these input pa-
rameter may be affecting the blocked shape of accelerograms as obtained for clays. A dependence of the
blocked shape on these parameters was indeed found. Generally lower ratio G0 / Gur and a higher value ϒ0.7
smoothen the blocked shape, which is a consequence of the less sudden changes of the system stiffness prop-
erties when the deviatoric stress rate tensor is changes sign. However in the complete range of reasonable
(empirically derived) values for G0 / Gur and ϒ0.7 the HSsmall model results the typical acceleration responses,
from which is concluded that the response sensitivity to these parameters is limited and they are not the criti-
cal factor in the unrealistic responses obtained.

The HSsmall model as initially applied to the soils includes stress dependency of stiffness. In order judge the ef-
fect of the resulting very low soil stiffness near seabed, the stress dependency of stiffness was removed by set-
ting stress dependency power m=0. Setting stress dependency power m=0 for the HSsmall and HS models rela-
tively most affects the top clay layer, since the lower limit stress dependent soil stiffness ratio:
actual soil stiffness c cos(ϕ )
= ≈ 0.05
reference soil stiffness c cos(ϕ ) + p ref sin(ϕ )
is not longer present in the model for very shallow soil material.

90
A comparison of the HSsmall, basic HS and the Mohr-Coulomb models was made in order to judge the effects
of the small strain overlay model combined with stress dependent stiffness development over depth. The re-
sults are shown in Figure 7.34 and Figure 7.35 for seabed level and the clays C1 and C2 layer interface for time
history interval 5-6 sec of the Duzce random signal.

Figure 7.34: Horizontal accelerations at seabed level, relation with soil model and stress dependency

Figure 7.35: Horizontal accelerations at C1-C2 interface, relation with soil model and stress dependency

As shown, the typical blocked behaviour was not obtained for the HS and MC model, that match very well for
overconsolidated soil, provided that Eur is taken equal to elastic Young’s modulus applied in MC model and
the stress dependency of stiffness artificially removed from the constitutive model. When the stress depend-
ency is included in the HS model, it gives accelerograms different from the accelerograms obtained for MC and
LE modelled soils. A phase shift is obtained and lower acceleration amplitudes result, which are directly a result
of the different stiffness. However, the sudden jumps in accelerations are not obtained. Figure 7.34 and Figure
7.35 also show the HSsmall soil response accelerograms where stress dependent stiffness is not included.
Compared to the stress dependent HSsmall and HS and MC soil models one may conclude:
- The unrealistic sudden acceleration development is far less pronounced when the HSsmall model stiff-
ness are made stress independent.
- At seabed level still some sudden changes in accelerations are obtained, but at the C1-C2 layer inter-
face a perfectly smooth response is obtained.

91
 - For the considered subinterval of the seismic event acceleration response amplitudes are converging
for the HSsmall, HS and Mohr-Coulomb soil models when stress dependent stiffness is not included.

For the initial analysis with HSsmall soil including stress dependent stiffness, the sudden developing accelera-
tions result almost ‘blocked’ accelerograms at seabed level, where at the C1-C2 layer interface this typically
shaped responses seem to be somewhat smoothened. At the C2-S3 layer interface the effect was not observed
at all, indicating the much stiffer sand layer to govern the local response at this depth. For the HSsmall soil
analysis without stress dependent stiffness, sudden changes obtained in accelerograms are less extreme at
seabed and not observed at all at the C1-C2 and C2-S3 layer interfaces.
These observations imply that the unrealistic shaped accelerograms are a result of the operation of HSsmall soil
model for soft soils in dynamics, were it seems to give most unrealistic results for the very soft material near
seabed. However the region where unrealistic responses are obtained seem to reach much deeper levels in the
clay layers.

Based on observations presented in the preceding, it seems that the conceptually very attractive HSsmall con-
stitutive model has its typical problems in dynamics, that are resulting from stress dependent very low stiffness
for top soil layers combined with the stiffness reset of the small strain overlay model. This reset of stiffness
may be realistic for static problems, but in dynamics stiffness development in reality needs a finite time inter-
val, which is not included in the HSsmall model and consequently results the unrealistic development of accel-
erations. However, the HSsmall features of including hysteretic damping, damping developing over time as a
function of strain amplitude, small strain stiffness and stress dependent stiffness, as obtained for real soils still
make the HSsmall model very attractive for dynamic geotechnical problems and consequently a solution for the
problem obtained is sought.
Considering possible solutions for the problem obtained and variations already studied, it is concluded that any
reasonable variations in HSsmall soil parameters for soft soils do not overcome the problem obtained. An at-
tempt was made to limit the problem by making only the top soil layer stiffness parameters stress independent
and keep the attractive stress dependent properties for deeper soil layers. For the top layer then averaged
stiffness parameters over depth are applied. The subinterval 0 – 12 sec. of the resulting acceleration response
at the C1-C2 layer interface is presented in Figure 7.36.

Figure 7.36: Horizontal accelerations at C1-C2 interface, HSsmall soil with only top layer stress independent stiffness

Although more realistic accelerograms are obtained compared to the models were all soil layers have stress
dependent stiffness parameters, again the unrealistic sudden changes in acceleration level are found in the re-
sponse for both clay layers. This observation implies that sudden changes obtained in accelerograms at greater
depths are not dominated by propagating waves generated near seabed, but may also result from local unreal-
istic HSsmall model behaviour.

92
7.2.3.3 Conclusions regarding soil modelling in dynamic analysis
Unrealistic (“blocked”) acceleration responses obtained by applying the HSsmall soil constitutive model were
found to be the direct result of stiffness reset at principal strain rate reversals. The effects of this stiffness up-
date seem to be a function of interaction of the small strain overlay model (with its input parameters) and the
stress dependency of stiffness can cause unrealistic accelerograms. Improvement of the HSsmall method for
dynamics is recommended, where implementation of a mobilisation time for stiffness reset probably is the so-
lution closest representing reality.

Consequently it might for this moment be more convenient to remove the stress dependent stiffness feature
from the HSsmall model for very weak soils. This can be done by artificially setting stress dependency power
m=0. By this means the attractive features of the HSsmall model related to hysteretic damping and small strain
stiffness development can be kept in the analyses. If required a soil profile may then be subdivided into more
layers, in order to overcome errors resulting from an oversimplification of the soil profile when stress depend-
ence of stiffness is not included anymore.

As an alternative linear elastic, Mohr-Coulomb or Hardening Soil constitutive models can be used for dynamic
analysis. However it is noted that Rayleigh damping as is to be applied in these models must be carefully se-
lected since it is by definition frequency dependent and the level is fixed for the total duration of the seismic
signal, irrespective of the motion intensity. Generally the viscous damping will be most accurate for saturated
permeable soils for which damping is governed by relative motions of pore water and soil particles. Damping
resulting from this phenomena is related to relative velocities and therefore relatively well described by viscous
damping. However, for dry or impermeable soils including viscous damping may be considered less appropri-
ate.

Mohr-Coulomb and Hardening Soil models include plasticity effects and hence will conceptually represent ac-
tual soil behaviour more accurate compared to linear elasticity. However, their disadvantages are the much
higher computational effort and no inclusion of hysteretic damping. Consequently in this study the dynamic
jetty analysis will be performed for the soil deposit modelled by HSsmall soil (with stress dependency power
m=0 for the top clay layer) and the simplified and computationally much more convenient linear elastic soil. To
this end Rayleigh damping added to the HSsmall soil will be assumed at 1-2% critical (paragraph 4.6.4.4) and
Rayleigh damping for the linear elastic soil will be estimated based on equivalent linear frequency domain
analysis as was done for the free field dynamic analysis.

7.2.4 Operation Plaxis 3D dynamics module (Release 2011.01 February 2012)

The operation of the new Plaxis 3D dynamic module was verified initially for the 3 Hz harmonic signal for sake
of simplicity and computational convenience, were linear elastic material behaviour with 5% critical added
Rayleigh viscous damping was applied. A plane strain configuration for a slice of soil was modelled in Plaxis 3D
by setting ymin and ymax boundaries to plane strain boundaries and xmin and xmax boundaries to default Plaxis vis-
cous boundaries. Figure 7.37 presents the seabed level acceleration and displacement responses for the Plaxis
2D and 3D models, which give almost identical results. It can be concluded that no significant 3D effects disturb
the free field response for bedrock motions applied in the x-direction of the x-z plane strain model configura-
tion.

93
 Figure 7.37: Plaxis 2D and Plaxis 3D site response for a 3 Hz harmonic bedrock excitation

Subsequently the free field responses due to the random Duzce bedrock signal obtained from the Plaxis 2D and
3D dynamic modules were compared. For computational convenience again linear elastic material was applied,
with Rayleigh damping ratios according to Table 7.3. Again a perfect fit for the Plaxis 2D and 3D modules was
found, providing confidence in proper operation of the very new Plaxis 3D dynamics module.

The good fit of the 2D and 3D dynamics modules provide good confidence in the performance of the new re-
leased 3D dynamics module and so allows to perform coupled soil deposit +jetty structure dynamic analysis
with Plaxis 3D. It is however noted that free field analysis with the HSsmall soil constitutive model is not per-
formed in Plaxis 3D because of time limitation, which however would be recommended for a future project.

7.3 Jetty simplified dynamic analysis according to Fajfar N2-response spectrum

method
The peak horizontal bedrock accelerations that are used as a starting point in the N2- response spectrum pro-
cedure are 0.7g and 1.02g for the earthquake intensities related to 10% and 2% probability of exceedance in 50
years respectively. applied at bedrock.

The resulting Turkish/ISO regulations response spectrum for the 1.02g soil deposit peak acceleration is shown
in Figure 7.38, where also the reduced response spectra for ductility levels 1.1, 1.5, 2.0, 3.0 and 5.0 are plotted
as an indication. The response spectrum shown below is plotted in the common acceleration-displacement re-
sponse spectrum (ADRS) -format, relating spectral accelerations to spectral pseudo-displacements by:
Su 1 T2
u (t ) = Ueiω0t → u&&(t ) = −ω0 2Ueiω0t → = 2 = 02 (7.4)
Su&& ω0 4π

94
 Figure 7.38: ISO normal response spectrum for bedrock acceleration 1.02g in ADRS format

It is noted that this ADRS representation is purely a tool in graphically representing the capacity and demand in
the same format, where it is not really contributing to the analytical solving procedure.
The key item in the simplified dynamic analysis is the nonlinear pushover curve, from which the fundamental
frequency for the linear range and consequently the load are estimated and the structure’s ductility capacity is
derived. The pushover curve is derived for lateral jetty analysis in Plaxis 3D subjected to an increasing lateral
load vector proportional to F = m ϕ aground , where m is the structures simplified mass vector and ϕ is the
normalized fundamental mode shape vector. The single point representation of the global structure pushover
characteristics is presented in Figure 7.39.

Figure 7.39: Single point representation of jetty global lateral load-deformation characteristic

A bilinear approximation of nonlinear global pushover characteristic is then made, from which initial stiffness
and equivalent mass the natural period of the equivalent SDOF system can be estimated to be T0 = 1.26 s .
Subsequently a coupling of seismic demand, reduction of demand by ductility related required displacement
demand is resulting the design global jetty displacement demand. This can be graphically represented in the

95
ADRS-format as shown in Figure 7.40 for the 1.02g earthquake with spectral shapes corresponding to deep
foundations according to ISO 19901:2004 – p14. For convenience the graphical representation of other load
levels and spectral configurations is not presented herein. The related results are summarized in Table 7.4 and
details can be found in appendix VIII.

Figure 7.40: Graphical representation of N2-method solution for seismic displacement demand

Table 7.4: Summary of required ductility and displacement demands for the different seismic load configurations
apeak bedrock Response spectrum characteristics Ductility demand μ Jetty deck dynamic dis-
placement demand
[g] [-] [m]
0.7 Soil class E, normal spectrum 0.96 (elastic range) 0.62
0.7 Soil class E, deep foundations spectrum 0.72 (elastic range) 0.47
1.02 Soil class E, normal spectrum 1.53 0.99
1.02 Soil class E, deep foundations spectrum 1.17 0.76

According to the pushover characteristics of the system now local response quantities can be determined,
which results pile head bending moments of 10200 and 11500 kNm for the 0.7g and 1.02g earthquake signal
respectively.

As noted before, single mode simplified dynamic approaches rely on the assumption that the structure re-
sponse can be assumed to be on the single fundamental mode. Literature and seismic standards generally rec-
ommend to take into account the number of modes up to a cumulative effective modal participating mass per-
centage of 90%. For the jetty structure transverse cross-section the modal masses where identified based on
eigenvalue analysis of the equivalent beam on Winkler foundation model in Seismostruct, with initial (lin-
earized) stiffness. The results are shown in Table 7.5.

Table 7.5: Jetty transverse cross-section modes and effective participating modal mass
Mode number fn Effective modal mass Cumulative effective modal mass
[-] [Hz] [-] [-]
1 0.84 0.556 0.556
2 4.14 0.167 0.724
3 5.78 0.051 0.775
4 8.62 0.130 0.905

96
The results show that considering the threshold value of 90%, the jetty structure response (see 4.5.3) may not
be assumed to be well represented by the fundamental mode only. Consequently the single mode simplified
dynamic analysis results might be somewhat non-conservative compared to results obtained from dynamic
analysis, which however will be strongly related to the frequency content of the seismic input signals adopted
in dynamic analysis. The sensitivity of obtained deformation demands to the included number of modes in sim-
plified dynamic analysis of jetty structures is recommended for further research.

7.4 Uncoupled and coupled site and jetty nonlinear dynamic time domain analysis
The transverse jetty responses from dynamic analysis are presented in this chapter. The focus herein is on the
jetty horizontal displacements at deck, seabed and pile tip level, the pile head bending moment development
of the outer right pile (pile 1), relative horizontal displacements of deck and seabed level to pile tip level (dy-
namic drift/residual drift), relative displacements of pile at seabed and surrounding soil at seabed (pile-soil in-
teraction) and the spectral representation of accelerations.
Both two different uncoupled dynamic analysis approaches and fully coupled dynamic analysis were per-
formed:
- Uncoupled dynamic analysis of structure in Seismostruct, were Winkler support node input signals are
determined from equivalent linear site response analysis
- Uncoupled dynamic analysis of structure in Seismostruct, were Winkler support node input signals are
determined from nonlinear Plaxis 2D site response analysis
- Coupled nonlinear dynamic analysis of soil deposit slice + jetty cross-section in Plaxis 3D
The system responses were calculated for Duzce and Kocaeli2 input signals scaled to 0.7g and 1.02g peak bed-
rock accelerations.

Additionally a single coupled 3D analysis was performed for the Kocaeli2 0.7g signal with a linear elastic soil in-
cluded. This analysis appeared to result responses almost identical to the uncoupled nonlinear structure +
equivalent linear soil analysis except from drifts resulting in permanent p-y elements. Because of its large com-
putational demand but limited added value it was not repeated for other signals and this approach is not fur-
ther taken into consideration.

The Plaxis 3D coupled analyses were performed for a 1.75 m slice of soil including three half jetty piles so ad-
vantage is taken of symmetry to limit the finite element model size. A model width of 300 m was adopted
which was sufficient with respect to boundary disturbances based on the free field model calibration. For more
details regarding the model configuration one is referred to paragraph

Dynamic analysis results are included in appendix IX, where this paragraph does provide a summary and the
important observed phenomena with explanatory figures.

7.4.1 Summary of acceleration, displacement and pile head bending moment response
quantities
Table 7.6 shows the extreme relative deck – pile tip displacements and pile head bending moments observed
from both uncoupled and coupled dynamic analysis during the earthquake.

97
Table 7.6: Dynamic peak displacements and pile head bending moments
Signal Response quantity Uncoupled analy- Uncoupled analy- Coupled analysis,
sis, equi. lin. soil sis, nonlinear soil nonlinear soil
Matlab + Seis- Plaxis 2D + Seis- Plaxis 3D
mostruct mostruct
2
Duzce 0.7g ax,min,deck [m/s ] -9.24 -7.52 -6.78
2
ax,max,deck [m/s ] 7.65 6.05 6.44
Δumin, deck-pile tip [m] -0.06 -0.17 -0.15
Δumax, deck-pile tip [m] 0.06 0.08 0.05
My,min,pile head [kNm] -3693 -3158 -2295
My,max,pile head [kNm] 3344 2775 2870
2
Duzce 1.02g ax,min,deck [m/s ] -11.32 -8.39 8,71
2
ax,max,deck [m/s ] 9.35 6.85 7.96
Δumin, deck-pile tip [m] -0.08 -0.17 -0.23
Δumax, deck-pile tip [m] 0.08 0.16 0.06
My,min,pile head [kNm] -4536 -3966 -2996
My,max,pile head [kNm] 4078 2915 3662
2
Kocaeli2 0.7g ax,min,deck [m/s ] -20.21 -9.05 -15.55
2
ax,max,deck [m/s ] 18.82 11.52 13.41
Δumin, deck-pile tip [m] -0.67 -0.64 -0.55
Δumax, deck-pile tip [m] 0.63 0.90 0.59
My,min,pile head [kNm] -11131 -9030 -6198
My,max,pile head [kNm] 11186 7440 6950
2
Kocaeli2 1.02g ax,min,deck [m/s ] -27.04 -13.43 -11.95
2
ax,max,deck [m/s ] 22.29 13.45 11.67
Δumin, deck-pile tip [m] -0.89 -1.08 -0.95
Δumax, deck-pile tip [m] 2.16 0.73 0.52
My,min,pile head [kNm] -13094 -7601 -7112
My,max,pile head [kNm] 11484 10317 8304

Dynamic response sensitivity to input signal selection can based on Table 7.6 be considered very high, since the
peak dynamic response quantities for the Kocaeli2 signal are by far exceeding the Duzce responses. The impor-
tance of considering a large number of seismic input signals is evident and seems to be a critical issue when dy-
namic analysis is used as a tool for design.

According to Table 7.6 one may conclude that with the uncoupled approach based on equivalent linear analysis
of the soil results the most extreme horizontal deck acceleration levels and the highest pile head bending mo-
ments. These higher response quantities are a direct consequence of higher soil deposit acceleration levels for
the equivalent linear soil in the highest intensity part of the seismic event. Figure 7.41 shows the much higher
energy content near the jetty fundamental frequency (0.8 Hz) for the case of equivalent linear soil by present-
ing the FFT of seabed soil accelerations. The coupled and uncoupled nonlinear approach show to be very simi-
lar, but reaching much lower peak spectral values, where the uncoupled analysis shows lower spectral values in
the high frequency range due to the stiffness proportional dashpot coefficients filtering out the higher fre-
quency content. These conclusions comply to all analysis performed (see appendix IX)

98
 Figure 7.41: FFT jetty pile at seabed horizontal acceleration spectra for Kocaeli2 0.7g input signal

According to Table 7.6 differential horizontal jetty deck – pile tip displacement extremes obtained over the sig-
nal time history are not systematically higher when one follows the uncoupled approach with equivalent linear
soil deposit analysis. However, as will be shown in the next paragraph dynamic displacement amplitudes ob-
tained from uncoupled analysis with equivalent linear soil analysis are typically higher where this is not true for
overall peak displacements, since no permanent deformations of the soil stratum are included when soil is as-
sumed to respond equivalent linear.

According to performance based design principles, the residual response quantities after an earthquake event
are also of importance, since they related to the post-earthquake serviceability performance of the structure.
Table 7.7 shows the residual relative horizontal displacements jetty deck – pile tip obtained from the per-
formed analysis.

Table 7.7: Residual displacements and pile head bending moments

Signal Response quantity Uncoupled analy- Uncoupled analy- Coupled analysis,
sis, equi. lin. soil sis, nonlinear soil nonlinear soil
Duzce 0.7g Δuresidual, deck-pile tip [m] 0 -0.12 -0.11
My,residual,pile head [kNm] -43 568 708
Duzce 1.02g Δuresidual, deck-pile tip [m] 0 -0.11 -0.18
My,residual,pile head [kNm] -51 294 879
Kocaeli2 0.7g Δuresidual, deck-pile tip [m] 0.07 0.55 0.25
My,residual,pile head [kNm] 521 -3231 -1440
Kocaeli2 1.02g Δuresidual, deck-pile tip [m] -0.26 0.39 0.21
My,residual,pile head [kNm] 1494 -2023 -1854

In relation to Table 7.7 it is noted that residual displacements of the system for the uncoupled approach based
on equivalent linear site response analysis are resulting from residual p-y spring displacements and residual pile
head moments/rotations only, where the other two approaches also allow permanent deformation of the soil
stratum to affect significantly the post-earthquake state of the system.

99
A remarkable observation from Table 7.7 can be made regarding the residual drifts resulting from the Kocaeli2
signal for the 0.7g and 1.02g intensities. The higher intensity signal namely is resulting lower residual drift. An
explanation was found from Figure 7.42 and Figure 7.43, that show the drift time histories for both intensities.
Apparently the higher intensity signal results development of permanent soil deformations and consequently
pile drifts fist in the negative direction and then in the positive direction, where the lower intensity signal
mainly develops permanent drift in the positive direction. Consequently the lower intensity signal according to
these analysis results a higher post-earthquake drift.

7.4.2 Displacement time histories and shifts of system equilibrium position

As was concluded in the previous paragraph, the conservatism of displacement level extremes cannot not di-
rectly be related to the analysis approach applied. This conclusion is supported by Figure 7.42 and Figure 7.43,
that show the differential jetty deck – pile tip horizontal displacements time histories obtained.

Figure 7.42: Relative horizontal displacement time history deck – pile tip, Kocaeli2 0.7g signal

Figure 7.43: Relative horizontal displacement time history deck-pile tip, Kocaeli2 1.02g signal

100
The displacement responses obtained from coupled and uncoupled analysis based on nonlinear soil deposit
analysis initially are corresponding very well. At the onset of very high shaking intensities a different in shift of
equilibrium position due to different global soil failure is obtained for the coupled and uncoupled nonlinear
analysis, which then becomes the new equilibrium position of the vibrating system. Then after development of
the different drifts again almost synchronous development of displacements is observed. The typical good cor-
respondence of coupled and uncoupled nonlinear analysis, expect from the sudden developments of different
drifts was found for all analysis performed (see appendix IX). The synchronous development of jetty displace-
ments for phases of less strong shaking provides a good confidence in the performance of the uncoupled
Winkler approach with nonlinear soil analysis for lower shaking levels without significant global soil failure.
The observed different drifts resulting after onset of global soil failure is explained by the fact that the piles
support the soil and hence reasonably a different permanent deformation will develop.

The figures clearly show how soil failure and related permanent soil stratum deformations may affect the jetty
displacements along the coupled analysis and uncoupled structure +nonlinear soil deposit analysis. This effect
was found to be strongly dependent on the input signal, and according to these figures even on the peak accel-
eration level to which the signal is scaled. Observations after past earthquakes confirm the residual drifts re-
sulting from soil failure to occur and design guidelines like PIANC impose limit values to these residual dis-
placements for designers. Consequently it can be concluded that simplified dynamic analysis procedures may
be useful for preliminary design stages but are not sufficient for final design stages as they do not consider re-
sidual displacements and related effects of global soil failure which typically is of importance because of the
deep foundation. The extreme sensitivity of the structure and soil deposit permanent deformations to the se-
lected seismic signal and intensity found in this study however make clear that dynamic analyses need to be
performed for a much larger number of input signals in order to be able to more reasonably estimate the ex-
pected permanent deformations. This conclusion is consistent with the generally larger required minimum
number of 7 (ISO recommendations, Turkish Technical Seismic Recommendations) and 3 (Eurocode) earth-
quake time history records.

Figure 7.42 and Figure 7.43 also clearly show the uncoupled approach based on equivalent linear analysis to re-
sult a much different development of displacements in time, with the strongest differences around onset of
global soil failure in the nonlinear models where this is not included for equivalent linear soil. Consequently
significantly different dynamic drifts are observed, that are not believed to be very realistic since based on the
nonlinear models the soil capacity mobilized to allow for these extreme motion levels is exceeds reasonable
soil strength.

For the uncoupled nonlinear analysis the soil deposit global plastic deformations are imposed to the structure
with Winkler foundation system. Then the structure responds to these imposed deformations, allowing for lo-
cal soil-structure interaction by p-y springs, but not affecting the development of the global soil deformations.
Here is the fundamental difference with the coupled analysis where global soil, local soil and structure re-
sponse are interacting. This interaction of jetty and soil becomes apparent from Figure 7.44 that clearly shows
how soil deposit displacements are affected by pile displacements.

101
Figure 7.44: Horizontal displacement contours, influence zone interaction of pile and soil, Kocaeli 1.02g signal at 18.00 s

The differences obtained in permanent structure drift and permanent soil deformations from the coupled and
uncoupled analysis probably are a consequence of this fundamental difference. However more future study re-
garding the pile, soil and interface displacement and acceleration developments in these strong motion phases
is required in order to investigate the precise effects of coupled/uncoupled interaction and Plaxis 3D interface
performance. The sensitivity of the jetty displacement response to variations in Winkler dashpot parameters of
the uncoupled model was found to be low. The development of gapping in Plaxis needs special attention and
its effect on the results is recommended for future study. The current Plaxis 3D interfaces are not accounting
for gap closure, resulting accumulating soil deformations surrounding the piles as shown in Figure 7.45.

Figure 7.45: Gapping of pile soil interface in Plaxis 3D

7.4.3 Kinematic pile loading

When one considers the bending moments in piles due to single inertial loading, the resulting pile deforma-
tions and bending moments develop as for inertial loading pushover analysis. However, in dynamic analysis the
kinematic pile loading contribution is included as well. Consequently bending moments and pile displacements
will not be coupled by the fundamental mode shape as was assumed in simplified dynamic analysis. Figure 7.46
shows the deformed pile shapes for four points in time for both coupled and uncoupled nonlinear analysis. The
results for both analyses show the same trend in pile deformation shapes, where for the uncoupled approach
higher drifts are observed. For the 5.6 s and 8.47 s point in time the soil deposit permanent deformations have

102
not developed and the deformation shape of the jetty is close to the fundamental mode shape. Kinematic pile
loading effects appear limited for these points in time since the first mode shape of the soft soil layers is not
pronounced in the pile deformation shape. This may be explained by the very soft soil conditions and the rela-
tively stiff pile. However for the 12.00 s and 36.00 s point in time the permanent soil deformations are clearly
present in the pile deformations as well. Hence may be concluded that for the typical situation considered with
stiff end-bearing piles in a soft soil deposit, kinematic pile loading by permanent soil deformations rather than
dynamic soil deformations strongly affect the jetty displacement response. It is noted that based on literature
the effects of dynamic kinematic pile loading will be more pronounced for layered soils with strongly varying
soil stiffness along the pile shaft.

Figure 7.46: Pile deformation shapes for Kocaeli2 1.02g signal at dynamic time 5.60 s, 8.47 s, 12.00 s and 36.00 s, left: re-
sults coupled analysis, right: results uncoupled analysis with nonlinear soil deposit analysis

The significance of kinematic pile loading when global soil failure is observed would be interesting for further
study. Especially the development of bending moments along the pile would be of interest,

7.4.4 Conclusions dynamic jetty analysis

Based on the preceding paragraphs it is concluded that:
- Coupled nonlinear analysis and uncoupled nonlinear analysis result reasonably synchronous responses
with respect to both horizontal displacements and pile head bending moments, except from a differ-
ence in jetty drift obtained after global soil failure. Whether this systematic difference in drift is di-
rectly related to the coupled vs. uncoupled approach, or if other model parameters as for example
Plaxis 3D interfaces might affect this behaviour is recommended for future study.
- From the analysis performed is becomes clear that global soil failure and related permanent structure
drifts may have a significant effect on the performance of the structure during and after an earth-
quake. This conclusion is confirmed by observations done after past earthquakes. From this perspec-
tive both the coupled analysis and the uncoupled analysis with nonlinear soil included are attractive
because a much better insight in effect of soil failure is obtained.
- Since the shifts in drift at global soil failure do not systematically result in higher deformation re-
sponses for either the coupled or the uncoupled dynamic analysis approach, not a single one can be
selected which is the most conservative. Investigation of their performance for a higher number of in-
put signals and different jetty structure and soil deposit configurations is recommended in order ob-
tain a better insight.
- At this moment Plaxis is implementing embedded pile elements that result 2.5D approach when ap-
plied to 2D plane strain soil models. To the authors best knowledge the performance of these embed-
ded piles for lateral pile loading however is still poor, which was the reason why they were not in-
cluded in this study. However when a better performance for lateral loading can be achieved embed-
ded pile elements probably may be very useful for practical dynamic soil-structure interaction prob-
lems.

103
7.5 Conclusions regarding jetty transverse seismic analysis

Table 7.8 and

Table 7.9 show the representative jetty response quantities for respectively the 0.7g and 1.02g level earth-
quake that were found for the different analysis approaches considered. The most conservative values ob-
tained from dynamic analysis for the Duzce and Kocaeli2 signals are included in these tables. Corresponding
minor damage and no-collapse performance requirements are also given as far as they are quantified in codes.

Table 7.8: Level 1 earthquake response

Design parame- Units Performance re- Simplified Uncoupled Uncoupled Coupled dy-
ter quirement dynamic dynamic dynamic namic analy-
analysis re- analysis, analysis, sis, non. lin
sults equi. lin soil non. lin soil soil
Plaxis 3D + Matlab + Plaxis 2D + Plaxis 3D
response Seismostruct Seismostruct
spectrum
procedure
Peak dynamic [m] According to client 0.62 0.67 0.90 0.59
displacement or serviceability
demand requirements of
installations lo-
cated at the jetty
Residual dis- [m] Pianc: res. tilt = 2° - 0.07 0.55 0.25
placement
ux,residual = 0.76m
Plastic hinge [-] 0.008 My,head=10200 My,head=11186 My,head=9030 My,head=6950
material strains kNm, kNm, kNm, kNm,
ε=0.0012< εy ε=0.011 ε=0.0011< εy ε=0.0008< εy

Table 7.9: Level 2 earthquake response

Design parame- Units Performance re- Simplified Uncoupled Uncoupled Coupled dy-
ter quirement dynamic dynamic dynamic namic analy-
analysis re- analysis, analysis, non. sis, non. lin.
sults equi. lin soil lin soil soil
Plaxis 3D + Matlab + Plaxis 2D + Plaxis 3D
response Seismostruct Seismostruct
spectrum
procedure
Peak dynamic [m] No no-collapse re- 0.99 2.16 1.08 0.95
displacement quirement jetty
and installations
on top of it like
safety of pipelines
for which re-
quirements differ
with varying codes
Residual dis- [m] - - 0.26 0.39 0.21
placement
Plastic hinge [-] 0.025 My,head=11500 My,head=13094 My,head=10317 My,head=8304
material strains kNm, kNm, ε=0.10 kNm, kNm,
ε=0.026 ε=0.0012< εy ε=0.0010< εy

Based on these results the following is noted:

- Responses obtained from simplified dynamic analysis are of the same order or magnitude as the re-
sponses from dynamic analysis and hence it seems reasonable to apply simplified dynamic analysis as
a tool in preliminary stages of seismic jetty design. However, it is noted that insight in actual earth-

104
 quake response and development of possible permanent deformations is not provided by this
method.
- The displacement responses obtained from these coupled/uncoupled nonlinear dynamic analyses
have shown to possibly exceed response levels obtained from simplified dynamic analysis, which
therefore cannot strictly be considered a conservative approach for design of jetty structures. More
advanced analysis seems to be required in final design stages.
- Including nonlinear soil in the dynamic soil-structure analysis limits pile head bending moments due to
high inertia loading, however the penalty is in the form of development of residual drifts as is also ob-
served after real earthquake events. Along modern performance based design principles this residual
drifts are also to be considered in design since they may be critical for post-earthquake serviceability
of the structure. From this perspective nonlinear dynamic analysis seems to be preferable, but prob-
lems are related to strong signal dependence which requires many signals to be taken into account.
- Coupled nonlinear analysis and uncoupled nonlinear analysis result reasonably synchronous responses
with respect to both horizontal displacements and pile head bending moments, except from a varying
difference in jetty drift obtained at global soil failure where after a new equilibrium position is found
and responses again are very similar. The synchronous responses imply the relatively dominating ef-
fect of pile tip excitation by the stiffer material, where the different residual drifts seem to be domi-
nated by the different soft soil deformations at failure for the coupled and uncoupled approach. The
latter difference in drift level is recommended for future studies in order to gain more confidence or
improve in the computationally very attractive uncoupled nonlinear dynamic analysis, which then
would be a very attractive tool fitting well in the concept of performance based design.
- Peak transverse horizontal displacements of the jetty as considered in this study show to reach con-
siderably high levels. When out of phase motions develop for different sections of the long jetty struc-
ture these relative horizontal displacements may even increase further, which may become critical
with respect to jetty deck capacity and safety of pipelines on top of the jetty. Uncoupled nonlinear
analysis that according to this study results reasonably accurate responses and is computationally very
efficient, may very well be used to study these possible directional effects on safety of jetty and instal-
lations.
- Simplified dynamic analysis was performed based on an peak bedrock excitation levels from PSHA.
When the same input acceleration levels were applied and equivalent linear site response analysis
output was used as input for uncoupled dynamic analysis, significantly higher and lower responses in
terms of drifts and pile head bending moments resulted for the Kocaeli2 and de Duzce signal respec-
tively. This implies the strong signal dependence, but reveals that the response spectrum procedures
according to the codes is not necessarily conservative in design. It is noted that the signals applied in
this study are recorded very near-fault signals, where the Kocaeli2 signal also is characterized by rela-
tively large bedrock displacements that then seems to strongly affect the deep pile foundation. The
frequency content of the signals close to the jetty fundamental frequency is not extremely high. It
seems therefore to be recommended for future projects to make sure that really the most unfavour-
able but realistic recorded accelerograms are accounted for in order to end up with a safe design.
Proper probabilistic seismic hazard assessment and carefully considering recorded past earthquakes of
the dominant fault(s) are of major importance herein.

105
8 Conclusions and Discussion

Performance based design and soil-structure interaction

Over the past decade the seismic design community has shown a clear trend switch towards performance
based design. Deformation quantities rather than inertia forces have become of main interest in verifying
earthquake resistance and related safety of structures. Besides, the effects of soil-foundation-structure interac-
tion have become increasingly important issues in seismic design. Along these observations three different de-
sign approaches for pile-deck systems in highly seismic areas were based on literature found to satisfactory, be-
ing simplified dynamic analysis (pushover + response spectrum), uncoupled and coupled nonlinear dynamic
analysis. The configuration of a jetty supported by large end-bearing piles in soft soil conditions subjected to
vertically propagating shear waves is considered, where it is noted that the same soil-structure interaction de-
sign concepts with some modifications can be applied to other pile supported onshore/offshore structures in
soft soil conditions.

Pushover analysis and p-y curves

A large series of pushover analysis was performed on single jetty piles and jetty pile groups, in order to com-
pare the performance of conventional Winkler p-y analysis to advanced nonlinear finite element analysis with
Plaxis 3D. Hardening soil parameter selection herein is found to be a critical issue, which has been addressed
through an extensive literature study. Based on single pile pushover analyses it was concluded that ISO p-y ex-
pressions for soft clays are very conservative, and applying them will not result realistic designs. Alternative ex-
pressions were sought in literature and based on comparison to Plaxis 3D hardening soil analysis the p-y ex-
pressions recently proposed by Jeanjean (2009) appear to perform very well. Subsequently pile group analysis
has shown pile efficiency reduction for side-side spaced piles with spacing less than 2.5*D and very significant
efficiency reduction for shadowing piles depending on the spacing. Plaxis 3D pushover analysis was in this
sense found to be a very valuable and practical tool for deriving efficiency factors for specific pile group con-
figurations.

Free field site response analysis

Free field site response analysis is an important step towards soil structure interaction modelling for jetties,
since the structure response is dominated by pile response and pile response is strongly affected by soil deposit
dynamic and permanent deformations. Free field responses from equivalent linear frequency domain analysis
and results from Plaxis 2D finite element analysis were compared. A good fit was found to linear elastic finite
element models. However, it was concluded that equivalent linear solutions may result significantly different
responses compared to nonlinear finite element analysis

Soil failure and permanent displacements

Post-earthquake investigations indicate that soil failure (plasticity) during seismic conditions causes permanent
displacements, which is often a critical factor for seismic performance of port structures. Hence nonlinear finite
element site response analysis is preferable since possible effects of soil failure can be better included in the
analysis. The Plaxis HSsmall soil constitutive model including hysteretic damping in the unload/reload range
and hardening plasticity is conceptually attractive from this perspective.

HSsmall performance
Verification of the model performance in dynamics in this study however has shown that the HSsmall model
needs improvements in order to overcome unrealistic behaviour in soft soil dynamics. The HSsmall sudden re-
set of stiffness at deviatoric strain rate reversals directly causes unrealistic sudden changes in acceleration lev-
els, where in reality stiffness mobilisation is believed to include a time factor smoothening the acceleration re-
sponse. Despite unrealistic acceleration levels resulting from HSsmall soft soil dynamics, displacement time his-
tories obtained appear reasonable, however they cannot be trusted and verification by analysis with different
soil models is required for the time being. Improvement of HSsmall model performance in site response analy-
sis was obtained when a constant stress independent stiffness of the top soil layer was assumed. However, a
reformulation of the HSsmall model seems to be required to make it suitable for dynamic analysis of soft soils.

106
Simplified dynamic analysis
Based on comparison of responses obtained from simplified dynamic, coupled and uncoupled dynamic design
approaches it was concluded that simplified dynamic analysis based on code provided spectra cannot be gen-
erally considered a conservative approach with respect to calculated displacement responses for jetty type
structures. This is explained by the importance of kinematic pile loading by the soil that in addition to super-
structure inertial loading may contribute significantly to the total response. Additionally it was found that re-
sponse spectra relating to structures with deep foundations did result significantly lower design responses for
the applied bedrock acceleration levels, which are by far exceeded by responses obtained from dynamic analy-
sis.

Coupled and uncoupled dynamic soil-structure analysis

Coupled soil structure interaction finite element analysis results very high computational demands and conse-
quently is not covering engineering design purposes. Alternatively uncoupled dynamic analysis of soil deposit
and jetty structure can be performed, which based on this study was concluded to be a promising approach for
engineering practice. This study has shown that very synchronous responses can be obtained from far less de-
manding uncoupled nonlinear analysis of site and structure, except from a different shift in jetty drift due to
different permanent soil deformations developing at global soil failure. The performance and possible im-
provements of the uncoupled approach in this sense need more study.

Recommended design approach

Based on the present state of knowledge the following recommendations for practicing engineers that need to
account for soil-structure interaction in jetty design are proposed:
- Plaxis 3D pushover analysis is an attractive tool to verify pushover characteristics of soil-foundation-
structure systems. Effects that typical soil profiles and pile group effects may have can be found from
Plaxis 3D pushover analysis and subsequently can be accounted for in equivalent Winkler model by de-
fining specific pile efficiency multipliers. The resulting validated Winker models may subsequently be
applied for dynamic analysis.
- Application of simplified dynamic analysis for preliminary design stages seems reasonable in order to
find basic dimensions of the structure. Soil acceleration levels may in this preliminary design stage
conservatively be estimated based on equivalent linear layered site response analysis for code speci-
fied bedrock acceleration levels.
- For final design stages uncoupled nonlinear analysis of soil deposit and structure seems to be the best
available tool for engineers to assess possible effects soil structure interaction in design. Since its
computational demand is limited it can be applied to find the dynamic response for a large number of
selected signals based on PSHA and by this means provides much more insight in possible jetty re-
sponse compared to simplified dynamic analysis. Additionally the low computational demand of the
uncoupled dynamic approach allows for considering 3D effects as for example torsional effects for jet-
ties and wharves constructed in slopes.

Finite element modelling

Finite element modelling has been an important part of this study, and some important conclusions regarding
the use of finite element analysis for engineering purposes have to be made because they have appeared to be
a critical factor in the design process. It is important to realize that as a designer you have the responsibility for
the models you build and the results that you get from them. Experience with the pre-processing certainly is
helps in this sense, but then still it was obtained that very extensive verification of all steps really is necessary in
order to make sure that output is reasonable. This also was found to be necessary for assumptions that initially
seem very obvious but may not be valid and introduce significant errors in the results. Additionally it has to be
concluded that for the time being computational demand required for 3D coupled nonlinear soil structure dy-
namic analysis exceeds reasonable limits for design purposes. “The sky is the limit” certainly still does not hold
for finite element software currently available to practising engineers, who consequently have to rely on
somehow simplified design techniques.

107
9 Suggestions for further research

Based on the present study the performance of the HSsmall soil constitutive model needs to be improved to
better describe soft soil dynamics. The introduction of a time factor in the reset of the HSsmall small strain
stiffness is considered to be a possible improvement since in real soils stiffness development also needs a finite
time interval which then will possibly smoothen the obtained response.

Based on the dynamic analysis output results the uncoupled dynamic analysis approach with nonlinear soil de-
posit analysis allowing for soil failure and permanent displacements included appears very promising for design
purposes. However further validation of the obtained differences in jetty drift resulting from soil permanent
deformations for the coupled and uncoupled analysis methods is recommended for future study. Verification
of the Plaxis 3D interface performance in dynamics is considered an important related issue.

In this study a single-mode simplified dynamic analysis method was adopted. However for building structures
multi-mode methods may increase accuracy of obtained results. Whether this may also be the case for jetty
structures is recommended for further study.

The conventional engineering assumption representing seismic input as vertically polarized shear waves was
adopted as a starting point in this study as these waves are often assumed to be critical. It is realized that this
may be an oversimplification, where Rayleigh waves and vertical motions may also affect the structure seismic
response. The effect Rayleigh waves and vertical soil motions may have on jetty structures hence are suggested
for future research.

The present study focuses on the case of soft clays overlying dense sand, both of which have a low liquefaction
potential. However when potentially liquefiable soils are present these definitely have to be accounted for in
design, since past earthquakes have shown the extreme effects laterally spreading soil layers may have on
structures with pile foundations embedded in these layers. Advance soil constitutive models allowing for pore
pressure build-up and liquefaction are nowadays becoming available, and also p-y expression including these
effects recently have been proposed (Bouckovalas, 2012) Validation of these continuum and Winkler models is
recommended for future study before one of them reasonably can be applied for design purposes.

108
Appendices

Since most appendices consist of Excel worksheets, Plaxis projects and Matlab scripts that are generally poorly
represented on A4 size hard-copy, it was decided to store them in digital format on the enclosed CD. A brief ta-
ble of contents is shown below, where the subdirectories when necessary contain a guiding document.

Appendix Contents
I Summary seismic design guidelines
II Soil parameter selection
III Winkler p-y curve expressions, dashpot coefficients and group efficiency reductions
IV Selected seismic bedrock signals
V Pushover analysis
VI Equivalent linear free field site response analysis
VII Finite element free field site response analysis
VIII Simplified dynamic jetty analysis
IX Uncoupled and coupled dynamic jetty analysis
References

- Allotey, N. K. & El Naggar, M. H. (2005). Cyclic Normal Force Displacement Model for Nonlinear Soil
Structure Interaction Analysis: Seismostruct implementation (Rep. No. GEOT 02-05). Geotechnical
Research Centre, Department of Civil & Environmental Engineering, University of Western Ontario,
London, Ontario, Canada.
- Alpan, I. (1970). Geotechnical Properties of Soils. Earth-Science Reviews, 6, 5-&.
- Aydinoglu, M. N. & Fahjan, Y. M. (2003). A unified formulation of the piecewise exact method for
inelastic seismic demand analysis including the P-delta effect. Earthquake Engineering & Structural
Dynamics, 32, 871-890.
- Barton, Y. O. (1982). Laterally Loaded Model Piles in Sand: Centrifuge Tests and Finite Element
Analyses. PhD Thesis University of Cambridge.
- Benz, T. (2006). Small-Strain Stiffness of Soils and its Numerical Consequences, PhD thesis. Universität
Stuttgart.
- Berger, B. S. (1977). Vibrations of An Infinite Orthotropic Layered Cylindrical Viscoelastic Shell in An
Acoustic Medium. Mechanical Engineering, 99, 105.
- Bouckovalas, G. (2012). Kinematic interaction of piles into laterally spreading soil. Proc.2nd
International conference on Performance Based Design in Earthquake Geotechnical Engineering.
- Boulanger, R. W., Curras, C. J., Kutter, B. L., Wilson, D. W., & Abghari, A. (1999). Seismic soil-pile-
structure interaction experiments and analyses. Journal of Geotechnical and Geoenvironmental
Engineering, 125, 750-759.
- Brinch Hansen J. (1961). The ultimate resistance of rigid piles against transversal forces. Danish
Geotechnical Institute, Copenhagen, Denmark, 98, 5-11.
- Brinkgreve, R. B. J., Kappert, M. H., & Bonnier, P. G. (2007). Hysteretic damping in a small-strain
stiffness model. Proc.NUMOG X, 737-742.
- Broms, B. B. (1964a). Lateral resistance of piles in cohesionless soils. Journal Soil Mechanics and
Foundations Division, ASCE.90(SM3), 123-156.
- Broms, B. B. (1964b). Lateral resistance of piles in cohesive soils. Journal Soil Mechanics and
Foundations Division, ASCE.90(SM2), 27-63.
- Chopra, A. K. (2001). Dynamics of Structures, Theory and Applications to Earthquake Engineering, 2nd
edition. Prentice Hall, New Jersey.
- Chopra, A. K. & Chintanapakdee, C. (2001). Drift Spectrum vs modal analysis of structural response to
near-fault ground motions. Earthquake Spectra, 17, 221-234.
- Chopra, A. K. & Goel, R. K. (2002). A modal pushover analysis procedure for estimating seismic
demands for buildings. Earthquake Engineering & Structural Dynamics, 31, 561-582.
- Chopra, A. K. & Goel, R. K. (2004). A modal pushover analysis procedure to estimate seismic demands
for unsymmetric-plan buildings. Earthquake Engineering & Structural Dynamics, 33, 903-927.
- Chopra, A. K., Goel, R. K., & Chintanapakdee, C. (2004). Evaluation of a modified MPA procedure
assuming higher modes as elastic to estimate seismic demands. Earthquake Spectra, 20, 757-778.
- Clough & Penzien (1993). Dynamics of Structures.
- DNV OS-J101 (2012).
- Dunnavant, T. W. & O'Neill, M. W. (1985). Performance, analysis and interpretation of a lateral load
test of a 72-inch-diameter bored pile in overconsolidated clay. Report UHCE 85-4, 57 pp..
- El Naggar, M. H. & Bentley, K. J. (2000). Dynamic analysis for laterally loaded piles and dynamic p-y
curves. Canadian Geotechnical Journal, 37, 1166-1183.
- El Naggar, M. H. a. N. M. (1996). Nonlinear analysis for dynamic lateral pile response. Soil Dynamics
and Earthquake Engineering, 15, 233-244.
- Erdik, M. e. al. (2004). Earthquake Hazard in Marmara Region, Turkey. Proc.13th World Conference of
Earthquake Engineering, Vancouver, Canada.
- Fajfar, P. (1999). Capacity spectrum method based on inelastic demand spectra. Earthquake
Engineering & Structural Dynamics, 28, 979-993.
- Freeman, S. A. (1998). Development and Use of Capacity Spectrum Method. In.
- Freeman, S. A. (2000). The Capacity Spectrum Method as a Tool for Seismic Design. In.
- Gazetas, G. (2012). Nonlinear Soil-Foundation-Structure Interaction. Proc.2nd International conference
on Performance Based Design in Earthquake Geotechnical Engineering.
- Gazetas, G. & Dobry, R. (1984a). Horizontal Response of Piles in Layered Soils. Journal of Geotechnical
Engineering-Asce, 110, 20-40.
- Gazetas, G. & Dobry, R. (1984b). Simple Radiation Damping Model for Piles and Footings. Journal of
Engineering Mechanics-Asce, 110, 937-956.
- Goel, R. K. (2010). Simplified Procedures for Seismic Analysis and Design of Piers and Wharves in
Marine Oil and LNG Terminals (Rep. No. California State Lands Commission, Contract No. C2005-051
and Dep. of the Navy, Office and Naval Research Award No. N00014-08-1-1-1209). California
Polytechnic State University, San Luis Obispo.
- Hanks, T. C. & Kanamori, H. (1979). Moment Magnitude Scale. Journal of Geophysical Research, 84,
2348-2350.
- Hardin, B. O. (1965). Dynamic Versus Static Shear Modulus for Dry Sand. Materials Research and
Standards, 5, 232-&.
- Hardin, B. O. & Black, W. L. (1969). Closure to vibration modulus of normally consolidated clays.
Proc.ASCE: Journal of the Soil Mechanics and Foundations Division, 98(SM7), 1531-1537.
- Hardin, B. O. & Drnevich, V. P. (1972). Shear Modulus and damping in soils. Proc.ASCE: Journal of the
Soil Mechanics and Foundations Division, 95(SM6), 1531-1537.
- Hartsuijker, C. W. J. W. (2004). Toegepaste Mechanica, Deel 3.
- Hashash, Y. M. A. & Park, D. (2002). Viscous damping formulation and high frequency motion
propagation in non-linear site response analysis. Soil Dynamics and Earthquake Engineering, 22, 611-
624.
- Idriss and Sun (1992). SHAKE91: a computer program for conducting equivalent linear seismic response
analyses of horizontally layered soil deposits, User's Guide University of California, Davis.
- Ishibashi, I. & Zhang, X. (1993). Unified Dynamic Shear Moduli and Damping Ratios of Sand and Clay.
Soils and Foudations, Japanese Society of Soil Mechanics and Foundation Engineering, 33, 182-191.
- ISO 19901-2, I. (2004). Petroleum and Natural Gas Industries, Specific requirements for offshore
structures - Part 2: Seismic design procedures and criteria International Organization for
Standardization,Technical Committee ISO/TC 67,Subcommittee SC 7.
- ISO 19902 (2006). Petroleum and Natural Gas Industries - Fixed Steel Offshore Structures International
Organization for Standardization,Technical Committee ISO/TC 67,Subcommittee SC 7.
- ISO 19902:2006 (2006). Petroleum and Natural Gas Industries - Fixed Steel Offshore Structures
International Organization for Standardization,Technical Committee ISO/TC 67,Subcommittee SC 7.
- ISO 19903 (2006). Petroleum and Natural Gas Industries, Fixed concrete offshore structures
International Organization of Standardization,Technical Committee ISO/TC67,Technical Committee
CEN/TC 1.
- Jeanjean, P. (2009). Re-assessment of P-Y Curves for Soft Clays from Centrifuge Testing and Finite
Element Modeling. Proc.Offshore Technology Conference (20158).
- Juirnarongrit, T. & Ashford, S. A. (2001). Effect of Pile Diameter on the Modulus of Sub-grade Reaction
(Rep. No. SSRP-2001/22). Department of Structural Engineering, University of California, San Diego.
- Kalkan, E. & Kunnath, S. K. (2006). Adaptive Modal Combination Procedure for Nonlinear Static
Analysis of Building Structures. Journal of Structural Engineering ASCE, 1721-1731.
- Kanamori, H. (1983). Magnitude Scale and Quantification of Earthquakes. Tectonophysics, 93, 185-199.
- Kavvadas, M. & Gazetas, G. (1993). Kinematic Seismic Response and Bending of Free-Head Piles in
Layered Soil. Geotechnique, 43, 207-222.
- Kramer, S. L. (1996). Geotechnical Earthquake Engineering.
- Kulhawy, F. H. & Mayne, P. W. (1990). Manual on Estimating Soil Properties for Foundation Design.
Cornell University ,Ithaca, New York.
- Lengkeek, H. J. (2003). Estimation of sand stiffness parameters from cone resistance. Plaxis Bulletin 13,
Januari 2002, 15-19.
- Lysmer, J. & Kuhlmeyer R.L. (1969). Finite Dynamic Model for Infinite Media. Journal of Engineering
and Mechnical Division, 859-877.
- Makris, N. & Gazetas, G. (1992). Dynamic Pile Soil Pile Interaction .2. Lateral and Seismic Response.
Earthquake Engineering & Structural Dynamics, 21, 145-162.
- Martin, C. M. & Randolph, M. F. (2006). Upper-bound analysis of lateral pile capacity in cohesive soil.
Geotechnique, 56, 141-145.
- Matlock, H. (1970). Correlations for design of laterally loaded piles in soft clay. Preprints Second
Annual Offshore Technology Conference, 1, 577-588.
- Matlock, H. (1979). Fugro Internal Memorandum. -.
- Meyerhof, G. G. (1995). Standard Penetration Tests and Pile Behavior Under Lateral Loads in
Cohesionless Soils. Canadian Geotechnical Journal, 32, 913-916.
- Michaels, P. (1998). In situ determination of soil stiffness and damping. Journal of Geotechnical and
Geoenvironmental Engineering, 124, 709-719.
- Michaels, P. (2006a). Comparison of Viscous Damping in Unsaturated Soils, Compression and Shear.
Unsaturated Soils, 2006, Vol.1, Proceedings of the Fourth International Conference on Unsaturated
Soils, (GSP 147), ASCE, Reston, VA.565-576.
- Michaels, P. (2006b). Relating damping to soil permeability. Int.Journal of Geomechanics, Vol.6 (3),
158-165.
- Michaels, P. (2008). Water, Inertial damping and the Complex shear modulus. ASCE.
- Miranda, E. & Bertero, V. V. (1994). Evaluation of Strength Reduction Factors for Earthquake-Resistant
Design. Earthquake Spectra, 10, 357-379.
- MOTEMS (2010). Marine Oil Terminal Engineering and Maintenance Standards, Chapter 31F of the
2010 California Building Code. California Building Standards Commission.
- Muir Wood, D. (1990). Soil Behaviour and Critical State Soil Mechanics. Cambridge University Press.
- Murff, J. D. & Hamilton, J. M. (1993). P-Ultimate for Undrained Analysis of Laterally Loaded Piles.
Journal of Geotechnical Engineering-Asce, 119, 91-107.
- Mylonakis, G. & Gazetas, G. (2000). Seismic soil-structure interaction: Beneficial or detrimental?
Journal of Earthquake Engineering, 4, 277-301.
- Mylonakis, G., Nikolaou, A., & Gazetas, G. (1997). Soil-pile-bridge seismic interaction: Kinematic and
inertial effects .1. Soft soil. Earthquake Engineering & Structural Dynamics, 26, 337-359.
- NEHRP Fema-450 (2012). NEHRP FEMA450.
- NEN-EN 1998 (2004). Eurocode NEN-EN 1998;2004 Design of structures for earthquake resistance.
European Committee for Standardization.
- Nogami, T., Otani, J., Konagai, K., & Chen, H. L. (1992). Nonlinear Soil-Pile Interaction-Model for
Dynamic Lateral Motion. Journal of Geotechnical Engineering-Asce, 118, 89-106.
- Novak, M. (1974). Dynamic stiffness and damping of piles. Canadian Geotechnical Journal, 11, 574-
598.
- O'Neill, M. W. & Dunnavant, T. W. (1984). A study of the effects of scale, velocity and cyclic
degradability on laterally loaded single piles in overconsolidated clay. Report UHCE 84-7: 368 pp..
- O'Neill, M. W. & Murchison, J. M. (1983). An evaluation of p-y relationships in sands. A report to the
American Petroleum Institute, PRAC 82-41-1.
- O'Neill, M. W., Reese, L. C., & Cox, W. R. (1990). Soil Behaviour for Piles Under Lateral Loading.
Proc.Offshore Technolgy Conference, Houston, Texas.
- PIANC, P. I. N. A. (2001). PIANC Design Guidelines for Port Structures. A.A. Balkema, Rotterdam,
Brookfield.
- Plaxis Material Models Manual (2011). Plaxis Material Models Manual, 2011 Plaxis b.v, Delft, The
Netherlands.
- Priestley, M. J. N., Calvi, G. M., & Kowalski, M. J. (2007). Displacement-Based Seismic Design of
Structures. IUSS Press, Pavia, Italy.
- Priestley, M. J. N. & Park, R. (1987). Strength and Ductility of Concrete Bridge Columns Under Seismic
Loading. Aci Structural Journal, 84, 61-76.
- Randolph, M. & Gouvernic, S. (2011). Offshore Geotechnical Engineering. Spon Press, Abingdon, Oxon.
- Randolph, M. F. & Houlsby, G. T. (1984). The limiting pressure on a circular pile loaded laterally in
cohesive soil. Geotechnique, 34(4), 613-623.
- Reese, L. C. & an Ympe, W. F. (2001). Single Piles and Pile Groups under Lateral Loading. A.A. Balkema,
Rotterdam, Brookfield.
- Reese, L. C., Cox, W. R., & Koop, F. D. (1974). Analysis of laterally loaded piles in sand. Proceedings
Fifth Annual Offshore Technology Conference, Houston TX.
- Reese, L. C. a. W. R. C. (1975). Lateral loading of deep foundations in stiff clay. Journal Geotechnical
Engineering Division, American Society of Civil Engineering, 101, 633-649.
- Roesset (1970). Fundamentals of Soil Amplification. Seismic Design for Nuclear Power Plants, 183-244.
- Santos, J. A. & Correia, A. G. (2001). Reference threshold shear strain of soil, its application to obtain
unique strain-dependent shear modulus curve for soil. Proc.15th International Conference on Soil
Mechanics and Geotechnical Engineering, Istanbul, Turkey, 1, 267-270.
- Schanz, T., Vermeer, P. A., & Bonnier, P. G. (1999). The Hardening-soil model: Formulation and
verification. Beyond 2000 in Computational Geotechnics, 281-290.
- Schnabel, P., Seed, H. B., & Lysmer, J. (1972). Modification of Seismograph Records for Effects of Local
Soil Conditions. Bulletin of the Seismological Society of America, 62, 1649-1664.
- Schonfield, A. N. & Wroth, P. (1968). Critical state soil mechanics. McGraw-Hill.
- Sigaran Loria, C. & Jaspers-Focks D.J (2011). HSS model adequacy in performance-based design
approach, Filyos New Port, Turkey. Proc.15th European Conference on Soil Mechanics and
Geotechnical Engineering, Istanbul, Turkey, 1579-1586.
- Sluys, L. J. & Borst de, R. (2010). Computational Methods in Nonlinear Solid Mechanics, Lecture Notes
CT5142. Delft University of Technology, The Netherlands.
- Stokoe et al, K. H. (2004). Development of a new family of normalized modulus reduction and material
damping curves. International Workshop on Uncertainties in Nonlinear Soil Properties and their Impact
on Modeling Dynamic Soil Response, UC Berkeley, CA.
- Stoll, R. (1985). Computer-aided studies of complex soil moduli. Proc., Measurement and Use of Shear
Wave Velocity for Evaluating Dynamic Soil Properties, ASCE, New York, 18-33.
- Transportation Research Board, N. R. C. (2001). NCHRP Report 461, Static and Dynamic Lateral Loading
of Pile Groups.
- Turkish Technical Seismic Regulations (2008). Turkish Technical Seismic Regulation on Construction of
Coastal and Harbour Structures, Railway and Airports. Turkish Ministry of Transportation.
- Verruijt, A. (2005). Soil Dynamics.
- Vesic (1961). Beam on elastic subgrade and the Winkler hypothesis. Proc.International Conference Soil
Mechanics and Foundation Engineering, Paris, 1, 845-850.
- Vucetic, M. & Dobry, R. (1991). Effect of Soil Plasticity on Cyclic Response. Journal of Geotechnical
Engineering-Asce, 117, 89-107.
- Wilson, J. F. (2003). Dynamics of Offshore Structures. John Wiley & Sons, Inc, Hoboken, New Jersey.
- Wolf, J. P. & Deeks, A. J. (2004). Foundation Vibration Analysis: A Strength-of-Materials Approach.
Elsevier, Oxford, U.K.